root / qemu-tech.texi @ feature-archipelago
History | View | Annotate | Download (22.9 kB)
1 | 1f673135 | bellard | \input texinfo @c -*- texinfo -*- |
---|---|---|---|
2 | debc7065 | bellard | @c %**start of header |
3 | debc7065 | bellard | @setfilename qemu-tech.info |
4 | e080e785 | Stefan Weil | |
5 | e080e785 | Stefan Weil | @documentlanguage en |
6 | e080e785 | Stefan Weil | @documentencoding UTF-8 |
7 | e080e785 | Stefan Weil | |
8 | debc7065 | bellard | @settitle QEMU Internals |
9 | debc7065 | bellard | @exampleindent 0 |
10 | debc7065 | bellard | @paragraphindent 0 |
11 | debc7065 | bellard | @c %**end of header |
12 | 1f673135 | bellard | |
13 | a1a32b05 | Stefan Weil | @ifinfo |
14 | a1a32b05 | Stefan Weil | @direntry |
15 | a1a32b05 | Stefan Weil | * QEMU Internals: (qemu-tech). The QEMU Emulator Internals. |
16 | a1a32b05 | Stefan Weil | @end direntry |
17 | a1a32b05 | Stefan Weil | @end ifinfo |
18 | a1a32b05 | Stefan Weil | |
19 | 1f673135 | bellard | @iftex |
20 | 1f673135 | bellard | @titlepage |
21 | 1f673135 | bellard | @sp 7 |
22 | 1f673135 | bellard | @center @titlefont{QEMU Internals} |
23 | 1f673135 | bellard | @sp 3 |
24 | 1f673135 | bellard | @end titlepage |
25 | 1f673135 | bellard | @end iftex |
26 | 1f673135 | bellard | |
27 | debc7065 | bellard | @ifnottex |
28 | debc7065 | bellard | @node Top |
29 | debc7065 | bellard | @top |
30 | debc7065 | bellard | |
31 | debc7065 | bellard | @menu |
32 | debc7065 | bellard | * Introduction:: |
33 | debc7065 | bellard | * QEMU Internals:: |
34 | debc7065 | bellard | * Regression Tests:: |
35 | debc7065 | bellard | * Index:: |
36 | debc7065 | bellard | @end menu |
37 | debc7065 | bellard | @end ifnottex |
38 | debc7065 | bellard | |
39 | debc7065 | bellard | @contents |
40 | debc7065 | bellard | |
41 | debc7065 | bellard | @node Introduction |
42 | 1f673135 | bellard | @chapter Introduction |
43 | 1f673135 | bellard | |
44 | debc7065 | bellard | @menu |
45 | 3aeaea65 | Max Filippov | * intro_features:: Features |
46 | 3aeaea65 | Max Filippov | * intro_x86_emulation:: x86 and x86-64 emulation |
47 | 3aeaea65 | Max Filippov | * intro_arm_emulation:: ARM emulation |
48 | 3aeaea65 | Max Filippov | * intro_mips_emulation:: MIPS emulation |
49 | 3aeaea65 | Max Filippov | * intro_ppc_emulation:: PowerPC emulation |
50 | 3aeaea65 | Max Filippov | * intro_sparc_emulation:: Sparc32 and Sparc64 emulation |
51 | 3aeaea65 | Max Filippov | * intro_xtensa_emulation:: Xtensa emulation |
52 | 3aeaea65 | Max Filippov | * intro_other_emulation:: Other CPU emulation |
53 | debc7065 | bellard | @end menu |
54 | debc7065 | bellard | |
55 | debc7065 | bellard | @node intro_features |
56 | 1f673135 | bellard | @section Features |
57 | 1f673135 | bellard | |
58 | 1f673135 | bellard | QEMU is a FAST! processor emulator using a portable dynamic |
59 | 1f673135 | bellard | translator. |
60 | 1f673135 | bellard | |
61 | 1f673135 | bellard | QEMU has two operating modes: |
62 | 1f673135 | bellard | |
63 | 1f673135 | bellard | @itemize @minus |
64 | 1f673135 | bellard | |
65 | 5fafdf24 | ths | @item |
66 | 998a0501 | blueswir1 | Full system emulation. In this mode (full platform virtualization), |
67 | 998a0501 | blueswir1 | QEMU emulates a full system (usually a PC), including a processor and |
68 | 998a0501 | blueswir1 | various peripherals. It can be used to launch several different |
69 | 998a0501 | blueswir1 | Operating Systems at once without rebooting the host machine or to |
70 | 998a0501 | blueswir1 | debug system code. |
71 | 1f673135 | bellard | |
72 | 5fafdf24 | ths | @item |
73 | 998a0501 | blueswir1 | User mode emulation. In this mode (application level virtualization), |
74 | 998a0501 | blueswir1 | QEMU can launch processes compiled for one CPU on another CPU, however |
75 | 998a0501 | blueswir1 | the Operating Systems must match. This can be used for example to ease |
76 | 998a0501 | blueswir1 | cross-compilation and cross-debugging. |
77 | 1f673135 | bellard | @end itemize |
78 | 1f673135 | bellard | |
79 | 1f673135 | bellard | As QEMU requires no host kernel driver to run, it is very safe and |
80 | 1f673135 | bellard | easy to use. |
81 | 1f673135 | bellard | |
82 | 1f673135 | bellard | QEMU generic features: |
83 | 1f673135 | bellard | |
84 | 5fafdf24 | ths | @itemize |
85 | 1f673135 | bellard | |
86 | 1f673135 | bellard | @item User space only or full system emulation. |
87 | 1f673135 | bellard | |
88 | debc7065 | bellard | @item Using dynamic translation to native code for reasonable speed. |
89 | 1f673135 | bellard | |
90 | 998a0501 | blueswir1 | @item |
91 | 998a0501 | blueswir1 | Working on x86, x86_64 and PowerPC32/64 hosts. Being tested on ARM, |
92 | 998a0501 | blueswir1 | HPPA, Sparc32 and Sparc64. Previous versions had some support for |
93 | 998a0501 | blueswir1 | Alpha and S390 hosts, but TCG (see below) doesn't support those yet. |
94 | 1f673135 | bellard | |
95 | 1f673135 | bellard | @item Self-modifying code support. |
96 | 1f673135 | bellard | |
97 | 1f673135 | bellard | @item Precise exceptions support. |
98 | 1f673135 | bellard | |
99 | 998a0501 | blueswir1 | @item |
100 | 998a0501 | blueswir1 | Floating point library supporting both full software emulation and |
101 | 998a0501 | blueswir1 | native host FPU instructions. |
102 | 998a0501 | blueswir1 | |
103 | 1f673135 | bellard | @end itemize |
104 | 1f673135 | bellard | |
105 | 1f673135 | bellard | QEMU user mode emulation features: |
106 | 5fafdf24 | ths | @itemize |
107 | 1f673135 | bellard | @item Generic Linux system call converter, including most ioctls. |
108 | 1f673135 | bellard | |
109 | 1f673135 | bellard | @item clone() emulation using native CPU clone() to use Linux scheduler for threads. |
110 | 1f673135 | bellard | |
111 | 5fafdf24 | ths | @item Accurate signal handling by remapping host signals to target signals. |
112 | 1f673135 | bellard | @end itemize |
113 | 1f673135 | bellard | |
114 | 998a0501 | blueswir1 | Linux user emulator (Linux host only) can be used to launch the Wine |
115 | 0adb1246 | Andreas Färber | Windows API emulator (@url{http://www.winehq.org}). A BSD user emulator for BSD |
116 | 998a0501 | blueswir1 | hosts is under development. It would also be possible to develop a |
117 | 998a0501 | blueswir1 | similar user emulator for Solaris. |
118 | 998a0501 | blueswir1 | |
119 | 1f673135 | bellard | QEMU full system emulation features: |
120 | 5fafdf24 | ths | @itemize |
121 | 998a0501 | blueswir1 | @item |
122 | 998a0501 | blueswir1 | QEMU uses a full software MMU for maximum portability. |
123 | 998a0501 | blueswir1 | |
124 | 998a0501 | blueswir1 | @item |
125 | 4a1418e0 | Anthony Liguori | QEMU can optionally use an in-kernel accelerator, like kvm. The accelerators |
126 | 4a1418e0 | Anthony Liguori | execute some of the guest code natively, while |
127 | 998a0501 | blueswir1 | continuing to emulate the rest of the machine. |
128 | 998a0501 | blueswir1 | |
129 | 998a0501 | blueswir1 | @item |
130 | 998a0501 | blueswir1 | Various hardware devices can be emulated and in some cases, host |
131 | 998a0501 | blueswir1 | devices (e.g. serial and parallel ports, USB, drives) can be used |
132 | 998a0501 | blueswir1 | transparently by the guest Operating System. Host device passthrough |
133 | 998a0501 | blueswir1 | can be used for talking to external physical peripherals (e.g. a |
134 | 998a0501 | blueswir1 | webcam, modem or tape drive). |
135 | 998a0501 | blueswir1 | |
136 | 998a0501 | blueswir1 | @item |
137 | 998a0501 | blueswir1 | Symmetric multiprocessing (SMP) even on a host with a single CPU. On a |
138 | 998a0501 | blueswir1 | SMP host system, QEMU can use only one CPU fully due to difficulty in |
139 | 998a0501 | blueswir1 | implementing atomic memory accesses efficiently. |
140 | 998a0501 | blueswir1 | |
141 | 1f673135 | bellard | @end itemize |
142 | 1f673135 | bellard | |
143 | debc7065 | bellard | @node intro_x86_emulation |
144 | 998a0501 | blueswir1 | @section x86 and x86-64 emulation |
145 | 1f673135 | bellard | |
146 | 1f673135 | bellard | QEMU x86 target features: |
147 | 1f673135 | bellard | |
148 | 5fafdf24 | ths | @itemize |
149 | 1f673135 | bellard | |
150 | 5fafdf24 | ths | @item The virtual x86 CPU supports 16 bit and 32 bit addressing with segmentation. |
151 | 998a0501 | blueswir1 | LDT/GDT and IDT are emulated. VM86 mode is also supported to run |
152 | 998a0501 | blueswir1 | DOSEMU. There is some support for MMX/3DNow!, SSE, SSE2, SSE3, SSSE3, |
153 | 998a0501 | blueswir1 | and SSE4 as well as x86-64 SVM. |
154 | 1f673135 | bellard | |
155 | 1f673135 | bellard | @item Support of host page sizes bigger than 4KB in user mode emulation. |
156 | 1f673135 | bellard | |
157 | 1f673135 | bellard | @item QEMU can emulate itself on x86. |
158 | 1f673135 | bellard | |
159 | 5fafdf24 | ths | @item An extensive Linux x86 CPU test program is included @file{tests/test-i386}. |
160 | 1f673135 | bellard | It can be used to test other x86 virtual CPUs. |
161 | 1f673135 | bellard | |
162 | 1f673135 | bellard | @end itemize |
163 | 1f673135 | bellard | |
164 | 1f673135 | bellard | Current QEMU limitations: |
165 | 1f673135 | bellard | |
166 | 5fafdf24 | ths | @itemize |
167 | 1f673135 | bellard | |
168 | 998a0501 | blueswir1 | @item Limited x86-64 support. |
169 | 1f673135 | bellard | |
170 | 1f673135 | bellard | @item IPC syscalls are missing. |
171 | 1f673135 | bellard | |
172 | 5fafdf24 | ths | @item The x86 segment limits and access rights are not tested at every |
173 | 1f673135 | bellard | memory access (yet). Hopefully, very few OSes seem to rely on that for |
174 | 1f673135 | bellard | normal use. |
175 | 1f673135 | bellard | |
176 | 1f673135 | bellard | @end itemize |
177 | 1f673135 | bellard | |
178 | debc7065 | bellard | @node intro_arm_emulation |
179 | 1f673135 | bellard | @section ARM emulation |
180 | 1f673135 | bellard | |
181 | 1f673135 | bellard | @itemize |
182 | 1f673135 | bellard | |
183 | 1f673135 | bellard | @item Full ARM 7 user emulation. |
184 | 1f673135 | bellard | |
185 | 1f673135 | bellard | @item NWFPE FPU support included in user Linux emulation. |
186 | 1f673135 | bellard | |
187 | 1f673135 | bellard | @item Can run most ARM Linux binaries. |
188 | 1f673135 | bellard | |
189 | 1f673135 | bellard | @end itemize |
190 | 1f673135 | bellard | |
191 | 24d4de45 | ths | @node intro_mips_emulation |
192 | 24d4de45 | ths | @section MIPS emulation |
193 | 24d4de45 | ths | |
194 | 24d4de45 | ths | @itemize |
195 | 24d4de45 | ths | |
196 | 24d4de45 | ths | @item The system emulation allows full MIPS32/MIPS64 Release 2 emulation, |
197 | 24d4de45 | ths | including privileged instructions, FPU and MMU, in both little and big |
198 | 24d4de45 | ths | endian modes. |
199 | 24d4de45 | ths | |
200 | 24d4de45 | ths | @item The Linux userland emulation can run many 32 bit MIPS Linux binaries. |
201 | 24d4de45 | ths | |
202 | 24d4de45 | ths | @end itemize |
203 | 24d4de45 | ths | |
204 | 24d4de45 | ths | Current QEMU limitations: |
205 | 24d4de45 | ths | |
206 | 24d4de45 | ths | @itemize |
207 | 24d4de45 | ths | |
208 | 24d4de45 | ths | @item Self-modifying code is not always handled correctly. |
209 | 24d4de45 | ths | |
210 | 24d4de45 | ths | @item 64 bit userland emulation is not implemented. |
211 | 24d4de45 | ths | |
212 | 24d4de45 | ths | @item The system emulation is not complete enough to run real firmware. |
213 | 24d4de45 | ths | |
214 | b1f45238 | ths | @item The watchpoint debug facility is not implemented. |
215 | b1f45238 | ths | |
216 | 24d4de45 | ths | @end itemize |
217 | 24d4de45 | ths | |
218 | debc7065 | bellard | @node intro_ppc_emulation |
219 | 1f673135 | bellard | @section PowerPC emulation |
220 | 1f673135 | bellard | |
221 | 1f673135 | bellard | @itemize |
222 | 1f673135 | bellard | |
223 | 5fafdf24 | ths | @item Full PowerPC 32 bit emulation, including privileged instructions, |
224 | 1f673135 | bellard | FPU and MMU. |
225 | 1f673135 | bellard | |
226 | 1f673135 | bellard | @item Can run most PowerPC Linux binaries. |
227 | 1f673135 | bellard | |
228 | 1f673135 | bellard | @end itemize |
229 | 1f673135 | bellard | |
230 | debc7065 | bellard | @node intro_sparc_emulation |
231 | 998a0501 | blueswir1 | @section Sparc32 and Sparc64 emulation |
232 | 1f673135 | bellard | |
233 | 1f673135 | bellard | @itemize |
234 | 1f673135 | bellard | |
235 | f6b647cd | blueswir1 | @item Full SPARC V8 emulation, including privileged |
236 | 3475187d | bellard | instructions, FPU and MMU. SPARC V9 emulation includes most privileged |
237 | a785e42e | blueswir1 | and VIS instructions, FPU and I/D MMU. Alignment is fully enforced. |
238 | 1f673135 | bellard | |
239 | a785e42e | blueswir1 | @item Can run most 32-bit SPARC Linux binaries, SPARC32PLUS Linux binaries and |
240 | a785e42e | blueswir1 | some 64-bit SPARC Linux binaries. |
241 | 3475187d | bellard | |
242 | 3475187d | bellard | @end itemize |
243 | 3475187d | bellard | |
244 | 3475187d | bellard | Current QEMU limitations: |
245 | 3475187d | bellard | |
246 | 5fafdf24 | ths | @itemize |
247 | 3475187d | bellard | |
248 | 3475187d | bellard | @item IPC syscalls are missing. |
249 | 3475187d | bellard | |
250 | 1f587329 | blueswir1 | @item Floating point exception support is buggy. |
251 | 3475187d | bellard | |
252 | 3475187d | bellard | @item Atomic instructions are not correctly implemented. |
253 | 3475187d | bellard | |
254 | 998a0501 | blueswir1 | @item There are still some problems with Sparc64 emulators. |
255 | 998a0501 | blueswir1 | |
256 | 998a0501 | blueswir1 | @end itemize |
257 | 998a0501 | blueswir1 | |
258 | 3aeaea65 | Max Filippov | @node intro_xtensa_emulation |
259 | 3aeaea65 | Max Filippov | @section Xtensa emulation |
260 | 3aeaea65 | Max Filippov | |
261 | 3aeaea65 | Max Filippov | @itemize |
262 | 3aeaea65 | Max Filippov | |
263 | 3aeaea65 | Max Filippov | @item Core Xtensa ISA emulation, including most options: code density, |
264 | 3aeaea65 | Max Filippov | loop, extended L32R, 16- and 32-bit multiplication, 32-bit division, |
265 | 044d003d | Max Filippov | MAC16, miscellaneous operations, boolean, FP coprocessor, coprocessor |
266 | 044d003d | Max Filippov | context, debug, multiprocessor synchronization, |
267 | 3aeaea65 | Max Filippov | conditional store, exceptions, relocatable vectors, unaligned exception, |
268 | 3aeaea65 | Max Filippov | interrupts (including high priority and timer), hardware alignment, |
269 | 3aeaea65 | Max Filippov | region protection, region translation, MMU, windowed registers, thread |
270 | 3aeaea65 | Max Filippov | pointer, processor ID. |
271 | 3aeaea65 | Max Filippov | |
272 | 044d003d | Max Filippov | @item Not implemented options: data/instruction cache (including cache |
273 | 044d003d | Max Filippov | prefetch and locking), XLMI, processor interface. Also options not |
274 | 044d003d | Max Filippov | covered by the core ISA (e.g. FLIX, wide branches) are not implemented. |
275 | 3aeaea65 | Max Filippov | |
276 | 3aeaea65 | Max Filippov | @item Can run most Xtensa Linux binaries. |
277 | 3aeaea65 | Max Filippov | |
278 | 3aeaea65 | Max Filippov | @item New core configuration that requires no additional instructions |
279 | 3aeaea65 | Max Filippov | may be created from overlay with minimal amount of hand-written code. |
280 | 3aeaea65 | Max Filippov | |
281 | 3aeaea65 | Max Filippov | @end itemize |
282 | 3aeaea65 | Max Filippov | |
283 | 998a0501 | blueswir1 | @node intro_other_emulation |
284 | 998a0501 | blueswir1 | @section Other CPU emulation |
285 | 1f673135 | bellard | |
286 | 998a0501 | blueswir1 | In addition to the above, QEMU supports emulation of other CPUs with |
287 | 998a0501 | blueswir1 | varying levels of success. These are: |
288 | 998a0501 | blueswir1 | |
289 | 998a0501 | blueswir1 | @itemize |
290 | 998a0501 | blueswir1 | |
291 | 998a0501 | blueswir1 | @item |
292 | 998a0501 | blueswir1 | Alpha |
293 | 998a0501 | blueswir1 | @item |
294 | 998a0501 | blueswir1 | CRIS |
295 | 998a0501 | blueswir1 | @item |
296 | 998a0501 | blueswir1 | M68k |
297 | 998a0501 | blueswir1 | @item |
298 | 998a0501 | blueswir1 | SH4 |
299 | 1f673135 | bellard | @end itemize |
300 | 1f673135 | bellard | |
301 | debc7065 | bellard | @node QEMU Internals |
302 | 1f673135 | bellard | @chapter QEMU Internals |
303 | 1f673135 | bellard | |
304 | debc7065 | bellard | @menu |
305 | debc7065 | bellard | * QEMU compared to other emulators:: |
306 | debc7065 | bellard | * Portable dynamic translation:: |
307 | debc7065 | bellard | * Condition code optimisations:: |
308 | debc7065 | bellard | * CPU state optimisations:: |
309 | debc7065 | bellard | * Translation cache:: |
310 | debc7065 | bellard | * Direct block chaining:: |
311 | debc7065 | bellard | * Self-modifying code and translated code invalidation:: |
312 | debc7065 | bellard | * Exception support:: |
313 | debc7065 | bellard | * MMU emulation:: |
314 | 998a0501 | blueswir1 | * Device emulation:: |
315 | debc7065 | bellard | * Hardware interrupts:: |
316 | debc7065 | bellard | * User emulation specific details:: |
317 | debc7065 | bellard | * Bibliography:: |
318 | debc7065 | bellard | @end menu |
319 | debc7065 | bellard | |
320 | debc7065 | bellard | @node QEMU compared to other emulators |
321 | 1f673135 | bellard | @section QEMU compared to other emulators |
322 | 1f673135 | bellard | |
323 | 1f673135 | bellard | Like bochs [3], QEMU emulates an x86 CPU. But QEMU is much faster than |
324 | 1f673135 | bellard | bochs as it uses dynamic compilation. Bochs is closely tied to x86 PC |
325 | 1f673135 | bellard | emulation while QEMU can emulate several processors. |
326 | 1f673135 | bellard | |
327 | 1f673135 | bellard | Like Valgrind [2], QEMU does user space emulation and dynamic |
328 | 1f673135 | bellard | translation. Valgrind is mainly a memory debugger while QEMU has no |
329 | 1f673135 | bellard | support for it (QEMU could be used to detect out of bound memory |
330 | 1f673135 | bellard | accesses as Valgrind, but it has no support to track uninitialised data |
331 | 1f673135 | bellard | as Valgrind does). The Valgrind dynamic translator generates better code |
332 | 1f673135 | bellard | than QEMU (in particular it does register allocation) but it is closely |
333 | 1f673135 | bellard | tied to an x86 host and target and has no support for precise exceptions |
334 | 1f673135 | bellard | and system emulation. |
335 | 1f673135 | bellard | |
336 | 1f673135 | bellard | EM86 [4] is the closest project to user space QEMU (and QEMU still uses |
337 | 1f673135 | bellard | some of its code, in particular the ELF file loader). EM86 was limited |
338 | 1f673135 | bellard | to an alpha host and used a proprietary and slow interpreter (the |
339 | 1f673135 | bellard | interpreter part of the FX!32 Digital Win32 code translator [5]). |
340 | 1f673135 | bellard | |
341 | 1f673135 | bellard | TWIN [6] is a Windows API emulator like Wine. It is less accurate than |
342 | 1f673135 | bellard | Wine but includes a protected mode x86 interpreter to launch x86 Windows |
343 | 36d54d15 | bellard | executables. Such an approach has greater potential because most of the |
344 | 1f673135 | bellard | Windows API is executed natively but it is far more difficult to develop |
345 | 1f673135 | bellard | because all the data structures and function parameters exchanged |
346 | 1f673135 | bellard | between the API and the x86 code must be converted. |
347 | 1f673135 | bellard | |
348 | 1f673135 | bellard | User mode Linux [7] was the only solution before QEMU to launch a |
349 | 1f673135 | bellard | Linux kernel as a process while not needing any host kernel |
350 | 1f673135 | bellard | patches. However, user mode Linux requires heavy kernel patches while |
351 | 1f673135 | bellard | QEMU accepts unpatched Linux kernels. The price to pay is that QEMU is |
352 | 1f673135 | bellard | slower. |
353 | 1f673135 | bellard | |
354 | 998a0501 | blueswir1 | The Plex86 [8] PC virtualizer is done in the same spirit as the now |
355 | 998a0501 | blueswir1 | obsolete qemu-fast system emulator. It requires a patched Linux kernel |
356 | 998a0501 | blueswir1 | to work (you cannot launch the same kernel on your PC), but the |
357 | 998a0501 | blueswir1 | patches are really small. As it is a PC virtualizer (no emulation is |
358 | 998a0501 | blueswir1 | done except for some privileged instructions), it has the potential of |
359 | 998a0501 | blueswir1 | being faster than QEMU. The downside is that a complicated (and |
360 | 998a0501 | blueswir1 | potentially unsafe) host kernel patch is needed. |
361 | 1f673135 | bellard | |
362 | 1f673135 | bellard | The commercial PC Virtualizers (VMWare [9], VirtualPC [10], TwoOStwo |
363 | 1f673135 | bellard | [11]) are faster than QEMU, but they all need specific, proprietary |
364 | 1f673135 | bellard | and potentially unsafe host drivers. Moreover, they are unable to |
365 | 1f673135 | bellard | provide cycle exact simulation as an emulator can. |
366 | 1f673135 | bellard | |
367 | 998a0501 | blueswir1 | VirtualBox [12], Xen [13] and KVM [14] are based on QEMU. QEMU-SystemC |
368 | 998a0501 | blueswir1 | [15] uses QEMU to simulate a system where some hardware devices are |
369 | 998a0501 | blueswir1 | developed in SystemC. |
370 | 998a0501 | blueswir1 | |
371 | debc7065 | bellard | @node Portable dynamic translation |
372 | 1f673135 | bellard | @section Portable dynamic translation |
373 | 1f673135 | bellard | |
374 | 1f673135 | bellard | QEMU is a dynamic translator. When it first encounters a piece of code, |
375 | 1f673135 | bellard | it converts it to the host instruction set. Usually dynamic translators |
376 | 1f673135 | bellard | are very complicated and highly CPU dependent. QEMU uses some tricks |
377 | 1f673135 | bellard | which make it relatively easily portable and simple while achieving good |
378 | 1f673135 | bellard | performances. |
379 | 1f673135 | bellard | |
380 | 998a0501 | blueswir1 | After the release of version 0.9.1, QEMU switched to a new method of |
381 | 998a0501 | blueswir1 | generating code, Tiny Code Generator or TCG. TCG relaxes the |
382 | 998a0501 | blueswir1 | dependency on the exact version of the compiler used. The basic idea |
383 | 998a0501 | blueswir1 | is to split every target instruction into a couple of RISC-like TCG |
384 | 998a0501 | blueswir1 | ops (see @code{target-i386/translate.c}). Some optimizations can be |
385 | 998a0501 | blueswir1 | performed at this stage, including liveness analysis and trivial |
386 | 998a0501 | blueswir1 | constant expression evaluation. TCG ops are then implemented in the |
387 | 998a0501 | blueswir1 | host CPU back end, also known as TCG target (see |
388 | 998a0501 | blueswir1 | @code{tcg/i386/tcg-target.c}). For more information, please take a |
389 | 998a0501 | blueswir1 | look at @code{tcg/README}. |
390 | 1f673135 | bellard | |
391 | debc7065 | bellard | @node Condition code optimisations |
392 | 1f673135 | bellard | @section Condition code optimisations |
393 | 1f673135 | bellard | |
394 | 998a0501 | blueswir1 | Lazy evaluation of CPU condition codes (@code{EFLAGS} register on x86) |
395 | 998a0501 | blueswir1 | is important for CPUs where every instruction sets the condition |
396 | 998a0501 | blueswir1 | codes. It tends to be less important on conventional RISC systems |
397 | f0f26a06 | Blue Swirl | where condition codes are only updated when explicitly requested. On |
398 | f0f26a06 | Blue Swirl | Sparc64, costly update of both 32 and 64 bit condition codes can be |
399 | f0f26a06 | Blue Swirl | avoided with lazy evaluation. |
400 | 998a0501 | blueswir1 | |
401 | 998a0501 | blueswir1 | Instead of computing the condition codes after each x86 instruction, |
402 | 998a0501 | blueswir1 | QEMU just stores one operand (called @code{CC_SRC}), the result |
403 | 998a0501 | blueswir1 | (called @code{CC_DST}) and the type of operation (called |
404 | 998a0501 | blueswir1 | @code{CC_OP}). When the condition codes are needed, the condition |
405 | 998a0501 | blueswir1 | codes can be calculated using this information. In addition, an |
406 | 998a0501 | blueswir1 | optimized calculation can be performed for some instruction types like |
407 | 998a0501 | blueswir1 | conditional branches. |
408 | 1f673135 | bellard | |
409 | 1235fc06 | ths | @code{CC_OP} is almost never explicitly set in the generated code |
410 | 1f673135 | bellard | because it is known at translation time. |
411 | 1f673135 | bellard | |
412 | f0f26a06 | Blue Swirl | The lazy condition code evaluation is used on x86, m68k, cris and |
413 | f0f26a06 | Blue Swirl | Sparc. ARM uses a simplified variant for the N and Z flags. |
414 | 1f673135 | bellard | |
415 | debc7065 | bellard | @node CPU state optimisations |
416 | 1f673135 | bellard | @section CPU state optimisations |
417 | 1f673135 | bellard | |
418 | 998a0501 | blueswir1 | The target CPUs have many internal states which change the way it |
419 | 998a0501 | blueswir1 | evaluates instructions. In order to achieve a good speed, the |
420 | 998a0501 | blueswir1 | translation phase considers that some state information of the virtual |
421 | 998a0501 | blueswir1 | CPU cannot change in it. The state is recorded in the Translation |
422 | 998a0501 | blueswir1 | Block (TB). If the state changes (e.g. privilege level), a new TB will |
423 | 998a0501 | blueswir1 | be generated and the previous TB won't be used anymore until the state |
424 | 998a0501 | blueswir1 | matches the state recorded in the previous TB. For example, if the SS, |
425 | 998a0501 | blueswir1 | DS and ES segments have a zero base, then the translator does not even |
426 | 998a0501 | blueswir1 | generate an addition for the segment base. |
427 | 1f673135 | bellard | |
428 | 1f673135 | bellard | [The FPU stack pointer register is not handled that way yet]. |
429 | 1f673135 | bellard | |
430 | debc7065 | bellard | @node Translation cache |
431 | 1f673135 | bellard | @section Translation cache |
432 | 1f673135 | bellard | |
433 | 27c8efcb | 陳韋任 | A 32 MByte cache holds the most recently used translations. For |
434 | 1f673135 | bellard | simplicity, it is completely flushed when it is full. A translation unit |
435 | 1f673135 | bellard | contains just a single basic block (a block of x86 instructions |
436 | 1f673135 | bellard | terminated by a jump or by a virtual CPU state change which the |
437 | 1f673135 | bellard | translator cannot deduce statically). |
438 | 1f673135 | bellard | |
439 | debc7065 | bellard | @node Direct block chaining |
440 | 1f673135 | bellard | @section Direct block chaining |
441 | 1f673135 | bellard | |
442 | 1f673135 | bellard | After each translated basic block is executed, QEMU uses the simulated |
443 | 1f673135 | bellard | Program Counter (PC) and other cpu state informations (such as the CS |
444 | 1f673135 | bellard | segment base value) to find the next basic block. |
445 | 1f673135 | bellard | |
446 | 1f673135 | bellard | In order to accelerate the most common cases where the new simulated PC |
447 | 1f673135 | bellard | is known, QEMU can patch a basic block so that it jumps directly to the |
448 | 1f673135 | bellard | next one. |
449 | 1f673135 | bellard | |
450 | 1f673135 | bellard | The most portable code uses an indirect jump. An indirect jump makes |
451 | 1f673135 | bellard | it easier to make the jump target modification atomic. On some host |
452 | 1f673135 | bellard | architectures (such as x86 or PowerPC), the @code{JUMP} opcode is |
453 | 1f673135 | bellard | directly patched so that the block chaining has no overhead. |
454 | 1f673135 | bellard | |
455 | debc7065 | bellard | @node Self-modifying code and translated code invalidation |
456 | 1f673135 | bellard | @section Self-modifying code and translated code invalidation |
457 | 1f673135 | bellard | |
458 | 1f673135 | bellard | Self-modifying code is a special challenge in x86 emulation because no |
459 | 1f673135 | bellard | instruction cache invalidation is signaled by the application when code |
460 | 1f673135 | bellard | is modified. |
461 | 1f673135 | bellard | |
462 | 1f673135 | bellard | When translated code is generated for a basic block, the corresponding |
463 | 998a0501 | blueswir1 | host page is write protected if it is not already read-only. Then, if |
464 | 998a0501 | blueswir1 | a write access is done to the page, Linux raises a SEGV signal. QEMU |
465 | 998a0501 | blueswir1 | then invalidates all the translated code in the page and enables write |
466 | 998a0501 | blueswir1 | accesses to the page. |
467 | 1f673135 | bellard | |
468 | 1f673135 | bellard | Correct translated code invalidation is done efficiently by maintaining |
469 | 1f673135 | bellard | a linked list of every translated block contained in a given page. Other |
470 | 5fafdf24 | ths | linked lists are also maintained to undo direct block chaining. |
471 | 1f673135 | bellard | |
472 | 998a0501 | blueswir1 | On RISC targets, correctly written software uses memory barriers and |
473 | 998a0501 | blueswir1 | cache flushes, so some of the protection above would not be |
474 | 998a0501 | blueswir1 | necessary. However, QEMU still requires that the generated code always |
475 | 998a0501 | blueswir1 | matches the target instructions in memory in order to handle |
476 | 998a0501 | blueswir1 | exceptions correctly. |
477 | 1f673135 | bellard | |
478 | debc7065 | bellard | @node Exception support |
479 | 1f673135 | bellard | @section Exception support |
480 | 1f673135 | bellard | |
481 | 1f673135 | bellard | longjmp() is used when an exception such as division by zero is |
482 | 5fafdf24 | ths | encountered. |
483 | 1f673135 | bellard | |
484 | 1f673135 | bellard | The host SIGSEGV and SIGBUS signal handlers are used to get invalid |
485 | 998a0501 | blueswir1 | memory accesses. The simulated program counter is found by |
486 | 998a0501 | blueswir1 | retranslating the corresponding basic block and by looking where the |
487 | 998a0501 | blueswir1 | host program counter was at the exception point. |
488 | 1f673135 | bellard | |
489 | 1f673135 | bellard | The virtual CPU cannot retrieve the exact @code{EFLAGS} register because |
490 | 1f673135 | bellard | in some cases it is not computed because of condition code |
491 | 1f673135 | bellard | optimisations. It is not a big concern because the emulated code can |
492 | 1f673135 | bellard | still be restarted in any cases. |
493 | 1f673135 | bellard | |
494 | debc7065 | bellard | @node MMU emulation |
495 | 1f673135 | bellard | @section MMU emulation |
496 | 1f673135 | bellard | |
497 | 998a0501 | blueswir1 | For system emulation QEMU supports a soft MMU. In that mode, the MMU |
498 | 998a0501 | blueswir1 | virtual to physical address translation is done at every memory |
499 | 998a0501 | blueswir1 | access. QEMU uses an address translation cache to speed up the |
500 | 998a0501 | blueswir1 | translation. |
501 | 1f673135 | bellard | |
502 | 1f673135 | bellard | In order to avoid flushing the translated code each time the MMU |
503 | 1f673135 | bellard | mappings change, QEMU uses a physically indexed translation cache. It |
504 | 5fafdf24 | ths | means that each basic block is indexed with its physical address. |
505 | 1f673135 | bellard | |
506 | 1f673135 | bellard | When MMU mappings change, only the chaining of the basic blocks is |
507 | 1f673135 | bellard | reset (i.e. a basic block can no longer jump directly to another one). |
508 | 1f673135 | bellard | |
509 | 998a0501 | blueswir1 | @node Device emulation |
510 | 998a0501 | blueswir1 | @section Device emulation |
511 | 998a0501 | blueswir1 | |
512 | 998a0501 | blueswir1 | Systems emulated by QEMU are organized by boards. At initialization |
513 | 998a0501 | blueswir1 | phase, each board instantiates a number of CPUs, devices, RAM and |
514 | 998a0501 | blueswir1 | ROM. Each device in turn can assign I/O ports or memory areas (for |
515 | 998a0501 | blueswir1 | MMIO) to its handlers. When the emulation starts, an access to the |
516 | 998a0501 | blueswir1 | ports or MMIO memory areas assigned to the device causes the |
517 | 998a0501 | blueswir1 | corresponding handler to be called. |
518 | 998a0501 | blueswir1 | |
519 | 998a0501 | blueswir1 | RAM and ROM are handled more optimally, only the offset to the host |
520 | 998a0501 | blueswir1 | memory needs to be added to the guest address. |
521 | 998a0501 | blueswir1 | |
522 | 998a0501 | blueswir1 | The video RAM of VGA and other display cards is special: it can be |
523 | 998a0501 | blueswir1 | read or written directly like RAM, but write accesses cause the memory |
524 | 998a0501 | blueswir1 | to be marked with VGA_DIRTY flag as well. |
525 | 998a0501 | blueswir1 | |
526 | 998a0501 | blueswir1 | QEMU supports some device classes like serial and parallel ports, USB, |
527 | 998a0501 | blueswir1 | drives and network devices, by providing APIs for easier connection to |
528 | 998a0501 | blueswir1 | the generic, higher level implementations. The API hides the |
529 | 998a0501 | blueswir1 | implementation details from the devices, like native device use or |
530 | 998a0501 | blueswir1 | advanced block device formats like QCOW. |
531 | 998a0501 | blueswir1 | |
532 | 998a0501 | blueswir1 | Usually the devices implement a reset method and register support for |
533 | 998a0501 | blueswir1 | saving and loading of the device state. The devices can also use |
534 | 998a0501 | blueswir1 | timers, especially together with the use of bottom halves (BHs). |
535 | 998a0501 | blueswir1 | |
536 | debc7065 | bellard | @node Hardware interrupts |
537 | 1f673135 | bellard | @section Hardware interrupts |
538 | 1f673135 | bellard | |
539 | e1b4382c | Stefan Weil | In order to be faster, QEMU does not check at every basic block if a |
540 | e8dc0938 | Stefan Weil | hardware interrupt is pending. Instead, the user must asynchronously |
541 | 1f673135 | bellard | call a specific function to tell that an interrupt is pending. This |
542 | 1f673135 | bellard | function resets the chaining of the currently executing basic |
543 | 1f673135 | bellard | block. It ensures that the execution will return soon in the main loop |
544 | 1f673135 | bellard | of the CPU emulator. Then the main loop can test if the interrupt is |
545 | 1f673135 | bellard | pending and handle it. |
546 | 1f673135 | bellard | |
547 | debc7065 | bellard | @node User emulation specific details |
548 | 1f673135 | bellard | @section User emulation specific details |
549 | 1f673135 | bellard | |
550 | 1f673135 | bellard | @subsection Linux system call translation |
551 | 1f673135 | bellard | |
552 | 1f673135 | bellard | QEMU includes a generic system call translator for Linux. It means that |
553 | 1f673135 | bellard | the parameters of the system calls can be converted to fix the |
554 | 1f673135 | bellard | endianness and 32/64 bit issues. The IOCTLs are converted with a generic |
555 | 1f673135 | bellard | type description system (see @file{ioctls.h} and @file{thunk.c}). |
556 | 1f673135 | bellard | |
557 | 1f673135 | bellard | QEMU supports host CPUs which have pages bigger than 4KB. It records all |
558 | 1f673135 | bellard | the mappings the process does and try to emulated the @code{mmap()} |
559 | 1f673135 | bellard | system calls in cases where the host @code{mmap()} call would fail |
560 | 1f673135 | bellard | because of bad page alignment. |
561 | 1f673135 | bellard | |
562 | 1f673135 | bellard | @subsection Linux signals |
563 | 1f673135 | bellard | |
564 | 1f673135 | bellard | Normal and real-time signals are queued along with their information |
565 | 1f673135 | bellard | (@code{siginfo_t}) as it is done in the Linux kernel. Then an interrupt |
566 | 1f673135 | bellard | request is done to the virtual CPU. When it is interrupted, one queued |
567 | 1f673135 | bellard | signal is handled by generating a stack frame in the virtual CPU as the |
568 | 1f673135 | bellard | Linux kernel does. The @code{sigreturn()} system call is emulated to return |
569 | 1f673135 | bellard | from the virtual signal handler. |
570 | 1f673135 | bellard | |
571 | 1f673135 | bellard | Some signals (such as SIGALRM) directly come from the host. Other |
572 | e8dc0938 | Stefan Weil | signals are synthesized from the virtual CPU exceptions such as SIGFPE |
573 | 1f673135 | bellard | when a division by zero is done (see @code{main.c:cpu_loop()}). |
574 | 1f673135 | bellard | |
575 | 1f673135 | bellard | The blocked signal mask is still handled by the host Linux kernel so |
576 | 1f673135 | bellard | that most signal system calls can be redirected directly to the host |
577 | 1f673135 | bellard | Linux kernel. Only the @code{sigaction()} and @code{sigreturn()} system |
578 | 1f673135 | bellard | calls need to be fully emulated (see @file{signal.c}). |
579 | 1f673135 | bellard | |
580 | 1f673135 | bellard | @subsection clone() system call and threads |
581 | 1f673135 | bellard | |
582 | 1f673135 | bellard | The Linux clone() system call is usually used to create a thread. QEMU |
583 | 1f673135 | bellard | uses the host clone() system call so that real host threads are created |
584 | 1f673135 | bellard | for each emulated thread. One virtual CPU instance is created for each |
585 | 1f673135 | bellard | thread. |
586 | 1f673135 | bellard | |
587 | 1f673135 | bellard | The virtual x86 CPU atomic operations are emulated with a global lock so |
588 | 1f673135 | bellard | that their semantic is preserved. |
589 | 1f673135 | bellard | |
590 | 1f673135 | bellard | Note that currently there are still some locking issues in QEMU. In |
591 | 1f673135 | bellard | particular, the translated cache flush is not protected yet against |
592 | 1f673135 | bellard | reentrancy. |
593 | 1f673135 | bellard | |
594 | 1f673135 | bellard | @subsection Self-virtualization |
595 | 1f673135 | bellard | |
596 | 1f673135 | bellard | QEMU was conceived so that ultimately it can emulate itself. Although |
597 | 1f673135 | bellard | it is not very useful, it is an important test to show the power of the |
598 | 1f673135 | bellard | emulator. |
599 | 1f673135 | bellard | |
600 | 1f673135 | bellard | Achieving self-virtualization is not easy because there may be address |
601 | 998a0501 | blueswir1 | space conflicts. QEMU user emulators solve this problem by being an |
602 | 998a0501 | blueswir1 | executable ELF shared object as the ld-linux.so ELF interpreter. That |
603 | 998a0501 | blueswir1 | way, it can be relocated at load time. |
604 | 1f673135 | bellard | |
605 | debc7065 | bellard | @node Bibliography |
606 | 1f673135 | bellard | @section Bibliography |
607 | 1f673135 | bellard | |
608 | 1f673135 | bellard | @table @asis |
609 | 1f673135 | bellard | |
610 | 5fafdf24 | ths | @item [1] |
611 | 1f673135 | bellard | @url{http://citeseer.nj.nec.com/piumarta98optimizing.html}, Optimizing |
612 | 1f673135 | bellard | direct threaded code by selective inlining (1998) by Ian Piumarta, Fabio |
613 | 1f673135 | bellard | Riccardi. |
614 | 1f673135 | bellard | |
615 | 1f673135 | bellard | @item [2] |
616 | 1f673135 | bellard | @url{http://developer.kde.org/~sewardj/}, Valgrind, an open-source |
617 | 1f673135 | bellard | memory debugger for x86-GNU/Linux, by Julian Seward. |
618 | 1f673135 | bellard | |
619 | 1f673135 | bellard | @item [3] |
620 | 1f673135 | bellard | @url{http://bochs.sourceforge.net/}, the Bochs IA-32 Emulator Project, |
621 | 1f673135 | bellard | by Kevin Lawton et al. |
622 | 1f673135 | bellard | |
623 | 1f673135 | bellard | @item [4] |
624 | 1f673135 | bellard | @url{http://www.cs.rose-hulman.edu/~donaldlf/em86/index.html}, the EM86 |
625 | 1f673135 | bellard | x86 emulator on Alpha-Linux. |
626 | 1f673135 | bellard | |
627 | 1f673135 | bellard | @item [5] |
628 | debc7065 | bellard | @url{http://www.usenix.org/publications/library/proceedings/usenix-nt97/@/full_papers/chernoff/chernoff.pdf}, |
629 | 1f673135 | bellard | DIGITAL FX!32: Running 32-Bit x86 Applications on Alpha NT, by Anton |
630 | 1f673135 | bellard | Chernoff and Ray Hookway. |
631 | 1f673135 | bellard | |
632 | 1f673135 | bellard | @item [6] |
633 | 1f673135 | bellard | @url{http://www.willows.com/}, Windows API library emulation from |
634 | 1f673135 | bellard | Willows Software. |
635 | 1f673135 | bellard | |
636 | 1f673135 | bellard | @item [7] |
637 | 5fafdf24 | ths | @url{http://user-mode-linux.sourceforge.net/}, |
638 | 1f673135 | bellard | The User-mode Linux Kernel. |
639 | 1f673135 | bellard | |
640 | 1f673135 | bellard | @item [8] |
641 | 5fafdf24 | ths | @url{http://www.plex86.org/}, |
642 | 1f673135 | bellard | The new Plex86 project. |
643 | 1f673135 | bellard | |
644 | 1f673135 | bellard | @item [9] |
645 | 5fafdf24 | ths | @url{http://www.vmware.com/}, |
646 | 1f673135 | bellard | The VMWare PC virtualizer. |
647 | 1f673135 | bellard | |
648 | 1f673135 | bellard | @item [10] |
649 | 5fafdf24 | ths | @url{http://www.microsoft.com/windowsxp/virtualpc/}, |
650 | 1f673135 | bellard | The VirtualPC PC virtualizer. |
651 | 1f673135 | bellard | |
652 | 1f673135 | bellard | @item [11] |
653 | 5fafdf24 | ths | @url{http://www.twoostwo.org/}, |
654 | 1f673135 | bellard | The TwoOStwo PC virtualizer. |
655 | 1f673135 | bellard | |
656 | 998a0501 | blueswir1 | @item [12] |
657 | 998a0501 | blueswir1 | @url{http://virtualbox.org/}, |
658 | 998a0501 | blueswir1 | The VirtualBox PC virtualizer. |
659 | 998a0501 | blueswir1 | |
660 | 998a0501 | blueswir1 | @item [13] |
661 | 998a0501 | blueswir1 | @url{http://www.xen.org/}, |
662 | 998a0501 | blueswir1 | The Xen hypervisor. |
663 | 998a0501 | blueswir1 | |
664 | 998a0501 | blueswir1 | @item [14] |
665 | 998a0501 | blueswir1 | @url{http://kvm.qumranet.com/kvmwiki/Front_Page}, |
666 | 998a0501 | blueswir1 | Kernel Based Virtual Machine (KVM). |
667 | 998a0501 | blueswir1 | |
668 | 998a0501 | blueswir1 | @item [15] |
669 | 998a0501 | blueswir1 | @url{http://www.greensocs.com/projects/QEMUSystemC}, |
670 | 998a0501 | blueswir1 | QEMU-SystemC, a hardware co-simulator. |
671 | 998a0501 | blueswir1 | |
672 | 1f673135 | bellard | @end table |
673 | 1f673135 | bellard | |
674 | debc7065 | bellard | @node Regression Tests |
675 | 1f673135 | bellard | @chapter Regression Tests |
676 | 1f673135 | bellard | |
677 | 1f673135 | bellard | In the directory @file{tests/}, various interesting testing programs |
678 | b1f45238 | ths | are available. They are used for regression testing. |
679 | 1f673135 | bellard | |
680 | debc7065 | bellard | @menu |
681 | debc7065 | bellard | * test-i386:: |
682 | debc7065 | bellard | * linux-test:: |
683 | debc7065 | bellard | @end menu |
684 | debc7065 | bellard | |
685 | debc7065 | bellard | @node test-i386 |
686 | 1f673135 | bellard | @section @file{test-i386} |
687 | 1f673135 | bellard | |
688 | 1f673135 | bellard | This program executes most of the 16 bit and 32 bit x86 instructions and |
689 | 1f673135 | bellard | generates a text output. It can be compared with the output obtained with |
690 | 1f673135 | bellard | a real CPU or another emulator. The target @code{make test} runs this |
691 | 1f673135 | bellard | program and a @code{diff} on the generated output. |
692 | 1f673135 | bellard | |
693 | 1f673135 | bellard | The Linux system call @code{modify_ldt()} is used to create x86 selectors |
694 | 1f673135 | bellard | to test some 16 bit addressing and 32 bit with segmentation cases. |
695 | 1f673135 | bellard | |
696 | 1f673135 | bellard | The Linux system call @code{vm86()} is used to test vm86 emulation. |
697 | 1f673135 | bellard | |
698 | 1f673135 | bellard | Various exceptions are raised to test most of the x86 user space |
699 | 1f673135 | bellard | exception reporting. |
700 | 1f673135 | bellard | |
701 | debc7065 | bellard | @node linux-test |
702 | 1f673135 | bellard | @section @file{linux-test} |
703 | 1f673135 | bellard | |
704 | 1f673135 | bellard | This program tests various Linux system calls. It is used to verify |
705 | 1f673135 | bellard | that the system call parameters are correctly converted between target |
706 | 1f673135 | bellard | and host CPUs. |
707 | 1f673135 | bellard | |
708 | debc7065 | bellard | @node Index |
709 | debc7065 | bellard | @chapter Index |
710 | debc7065 | bellard | @printindex cp |
711 | debc7065 | bellard | |
712 | debc7065 | bellard | @bye |