diff options
author | bellard <bellard@c046a42c-6fe2-441c-8c8c-71466251a162> | 2005-02-10 21:46:47 +0000 |
---|---|---|
committer | bellard <bellard@c046a42c-6fe2-441c-8c8c-71466251a162> | 2005-02-10 21:46:47 +0000 |
commit | 7c3fc84d86764fddd8dbc5408409eb5b2b4dc96e (patch) | |
tree | ea8a7d63be436893c00d5af87e75b69289e4f22f /qemu-doc.texi | |
parent | d7ce296f57740c076279fd6a0282bea255ec0842 (diff) |
update
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@1276 c046a42c-6fe2-441c-8c8c-71466251a162
Diffstat (limited to 'qemu-doc.texi')
-rw-r--r-- | qemu-doc.texi | 272 |
1 files changed, 153 insertions, 119 deletions
diff --git a/qemu-doc.texi b/qemu-doc.texi index ca5cb09b05..9c25728b6a 100644 --- a/qemu-doc.texi +++ b/qemu-doc.texi @@ -34,8 +34,12 @@ to ease cross-compilation and cross-debugging. @end itemize -As QEMU requires no host kernel driver to run, it is very safe and -easy to use. +QEMU can run without an host kernel driver and yet gives acceptable +performance. On an x86 host, if you want the highest performance for +the x86 target, the @emph{QEMU Accelerator Module} (KQEMU) allows QEMU +to reach near native performances. KQEMU is currently only supported +for an x86 Linux 2.4 or 2.6 host system, but more host OSes will be +supported in the future. For system emulation, the following hardware targets are supported: @itemize @@ -52,14 +56,8 @@ If you want to compile QEMU yourself, see @ref{compilation}. @section Linux -Download the binary distribution (@file{qemu-XXX-i386.tar.gz}) in -@file{/tmp} and untar it as root from @file{/}: - -@example -su -cd / -tar zxvf /tmp/qemu-XXX-i386.tar.gz -@end example +If a precompiled package is available for your distribution - you just +have to install it. Otherwise, see @ref{compilation}. @section Windows @@ -77,33 +75,8 @@ Download the experimental binary installer at @c man begin DESCRIPTION -The QEMU System emulator simulates a complete PC. - -In order to meet specific user needs, two versions of QEMU are -available: - -@enumerate - -@item -@code{qemu-fast} uses the host Memory Management Unit (MMU) to -simulate the x86 MMU. It is @emph{fast} but has limitations because -the whole 4 GB address space cannot be used and some memory mapped -peripherials cannot be emulated accurately yet. Therefore, a specific -guest Linux kernel can be used (@xref{linux_compile}) as guest -OS. - -Moreover there is no separation between the host and target address -spaces, so it offers no security (the target OS can modify the -@code{qemu-fast} code by writing at the right addresses). - -@item -@code{qemu} uses a software MMU. It is about @emph{two times slower} -but gives a more accurate emulation and a complete separation between -the host and target address spaces. - -@end enumerate - -QEMU emulates the following PC peripherials: +The QEMU System emulator simulates the +following PC peripherials: @itemize @minus @item @@ -369,6 +342,12 @@ Force hard disk 0 physical geometry (1 <= @var{c} <= 16383, 1 <= translation mode (@var{t}=none, lba or auto). Usually QEMU can guess all thoses parameters. This option is useful for old MS-DOS disk images. + +@item -no-kqemu +Disable the usage of the QEMU Accelerator module (KQEMU). QEMU will work as +usual but will be slower. This option can be useful to determine if +emulation problems are coming from KQEMU. + @item -isa Simulate an ISA-only system (default is PCI system). @item -std-vga @@ -444,6 +423,25 @@ Fabrice Bellard @end ignore +@section QEMU Accelerator Module + +The QEMU Accelerator Module (KQEMU) is an optional part of QEMU currently only +available for Linux 2.4 or 2.6 x86 hosts. It enables QEMU to run x86 +code much faster. Provided it is installed on your PC (see +@ref{kqemu_install}), QEMU will automatically use it. + +WARNING: as with any alpha stage kernel driver, KQEMU may cause +arbitrary data loss on your PC, so you'd better backup your sensitive +data before using it. + +When using KQEMU, QEMU will create a big hidden file containing the +RAM of the virtual machine. For best performance, it is important that +this file is kept in RAM and not on the hard disk. QEMU uses the +@file{/dev/shm} directory to create this file because @code{tmpfs} is +usually mounted on it (check with the shell command +@code{df}). Otherwise @file{/tmp} is used as fallback. You can use the +@var{QEMU_TMPDIR} shell variable to set a new directory for the QEMU +RAM file. @section QEMU Monitor @@ -825,12 +823,6 @@ A 2.5.74 kernel is also included in the archive. Just replace the bzImage in qemu.sh to try it. @item -qemu-fast creates a temporary file in @var{$QEMU_TMPDIR} (@file{/tmp} is the -default) containing all the simulated PC memory. If possible, try to use -a temporary directory using the tmpfs filesystem to avoid too many -unnecessary disk accesses. - -@item In order to exit cleanly from qemu, you can do a @emph{shutdown} inside qemu. qemu will automatically exit when the Linux shutdown is done. @@ -848,80 +840,6 @@ Lawton for the plex86 Project (@url{www.plex86.org}). @end enumerate -@node linux_compile -@section Linux Kernel Compilation - -You can use any linux kernel with QEMU. However, if you want to use -@code{qemu-fast} to get maximum performances, you must use a modified -guest kernel. If you are using a 2.6 guest kernel, you can use -directly the patch @file{linux-2.6-qemu-fast.patch} made by Rusty -Russel available in the QEMU source archive. Otherwise, you can make the -following changes @emph{by hand} to the Linux kernel: - -@enumerate -@item -The kernel must be mapped at 0x90000000 (the default is -0xc0000000). You must modify only two lines in the kernel source: - -In @file{include/asm/page.h}, replace -@example -#define __PAGE_OFFSET (0xc0000000) -@end example -by -@example -#define __PAGE_OFFSET (0x90000000) -@end example - -And in @file{arch/i386/vmlinux.lds}, replace -@example - . = 0xc0000000 + 0x100000; -@end example -by -@example - . = 0x90000000 + 0x100000; -@end example - -@item -If you want to enable SMP (Symmetric Multi-Processing) support, you -must make the following change in @file{include/asm/fixmap.h}. Replace -@example -#define FIXADDR_TOP (0xffffX000UL) -@end example -by -@example -#define FIXADDR_TOP (0xa7ffX000UL) -@end example -(X is 'e' or 'f' depending on the kernel version). Although you can -use an SMP kernel with QEMU, it only supports one CPU. - -@item -If you are not using a 2.6 kernel as host kernel but if you use a target -2.6 kernel, you must also ensure that the 'HZ' define is set to 100 -(1000 is the default) as QEMU cannot currently emulate timers at -frequencies greater than 100 Hz on host Linux systems < 2.6. In -@file{include/asm/param.h}, replace: - -@example -# define HZ 1000 /* Internal kernel timer frequency */ -@end example -by -@example -# define HZ 100 /* Internal kernel timer frequency */ -@end example - -@end enumerate - -The file config-2.x.x gives the configuration of the example kernels. - -Just type -@example -make bzImage -@end example - -As you would do to make a real kernel. Then you can use with QEMU -exactly the same kernel as you would boot on your PC (in -@file{arch/i386/boot/bzImage}). - @node gdb_usage @section GDB usage @@ -976,6 +894,12 @@ When using a 2.6 guest Linux kernel, you should add the option kernels make very strict real time clock checks by default that QEMU cannot simulate exactly. +When using a 2.6 guest Linux kernel, verify that the 4G/4G patch is +not activated because QEMU is slower with this patch. The QEMU +Accelerator Module is also much slower in this case. Earlier Fedora +Core 3 Linux kernel (< 2.6.9-1.724_FC3) were known to incorporte this +patch by default. Newer kernels don't have it. + @subsection Windows If you have a slow host, using Windows 95 is better as it gives the @@ -1226,9 +1150,119 @@ Act as if the host page size was 'pagesize' bytes @node compilation @chapter Compilation from the sources -@section Linux/BSD +@section Linux/Unix + +@subsection Compilation + +First you must decompress the sources: +@example +cd /tmp +tar zxvf qemu-x.y.z.tar.gz +cd qemu-x.y.z +@end example + +Then you configure QEMU and build it (usually no options are needed): +@example +./configure +make +@end example + +Then type as root user: +@example +make install +@end example +to install QEMU in @file{/usr/local}. + +@node kqemu_install +@subsection QEMU Accelerator Installation + +If you use x86 Linux, the compilation of the QEMU Accelerator Kernel +Module (KQEMU) is automatically activated provided you have the +necessary kernel headers. If nonetheless the compilation fails, you +can disable its compilation with the @option{--disable-kqemu} option. + +If you are using a 2.6 host kernel, then all the necessary kernel +headers should be already installed. If you are using a 2.4 kernel, +then you should verify that properly configured kernel sources are +installed and compiled. On a Redhat 9 distribution for example, the +following must be done: +@example +1) Install the kernel-source-xxx package +2) cd /usr/src/linux-xxx +3) make distclean +4) Copy /boot/config-vvv in .config (use uname -r to know your configuration name 'vvv') +5) Edit the Makefile to change the EXTRAVERSION line to match your + current configuration name: + EXTRAVERSION = -custom +to + EXTRAVERSION = -8 # This is an example, it can be -8smp too +5) make menuconfig # Just save the configuration +6) make dep bzImage +@end example -Read the @file{README} which gives the related information. +The installation of KQEMU is not fully automatic because it is highly +distribution dependent. When launching +@example +make install +@end example + +KQEMU is installed in /lib/modules/@var{kernel_version}/misc. The +device @file{/dev/kqemu} is created with read/write access rights for +everyone. If you fear security issues, you can restrict the access +rights of @file{/dev/kqemu}. + +If you want that KQEMU is installed automatically at boot time, you can add + +@example +# Load the KQEMU kernel module +/sbin/modprobe kqemu +@end example + +in @file{/etc/rc.d/rc.local}. + +If your distribution uses udev (like Fedora), the @file{/dev/kqemu} is +not created automatically (yet) at every reboot. You can add the +following in @file{/etc/rc.d/rc.local}: + +@example +# Create the KQEMU device +mknod /dev/kqemu c 254 0 +chmod 666 /dev/kqemu +@end example + +@subsection Tested tool versions + +In order to compile QEMU succesfully, it is very important that you +have the right tools. The most important one is gcc. I cannot guaranty +that QEMU works if you do not use a tested gcc version. Look at +'configure' and 'Makefile' if you want to make a different gcc +version work. + +@example +host gcc binutils glibc linux distribution +---------------------------------------------------------------------- +x86 3.2 2.13.2 2.1.3 2.4.18 + 2.96 2.11.93.0.2 2.2.5 2.4.18 Red Hat 7.3 + 3.2.2 2.13.90.0.18 2.3.2 2.4.20 Red Hat 9 + +PowerPC 3.3 [4] 2.13.90.0.18 2.3.1 2.4.20briq + 3.2 + +Alpha 3.3 [1] 2.14.90.0.4 2.2.5 2.2.20 [2] Debian 3.0 + +Sparc32 2.95.4 2.12.90.0.1 2.2.5 2.4.18 Debian 3.0 + +ARM 2.95.4 2.12.90.0.1 2.2.5 2.4.9 [3] Debian 3.0 + +[1] On Alpha, QEMU needs the gcc 'visibility' attribute only available + for gcc version >= 3.3. +[2] Linux >= 2.4.20 is necessary for precise exception support + (untested). +[3] 2.4.9-ac10-rmk2-np1-cerf2 + +[4] gcc 2.95.x generates invalid code when using too many register +variables. You must use gcc 3.x on PowerPC. +@end example @section Windows |