[Qemu-devel] qemu on 2.6.0-rc1

[Qemu-devel] qemu on 2.6.0-rc1

[Rutger Nijlunsing]

Hi,

I try to run qemu-fast on Linux kernel 2.6.0-rc1. I could not find the
things which I should change in the documentation:

    # The kernel must be mapped at 0x90000000 (the default is 0xc0000000). You must modify only two lines in the kernel source: In `include/asm/page.h', replace
    #define __PAGE_OFFSET           (0xc0000000)
    by
    #define __PAGE_OFFSET           (0x90000000)
    
    And in `arch/i386/vmlinux.lds', replace
      . = 0xc0000000 + 0x100000;
    by
      . = 0x90000000 + 0x100000;

The problems are that vmlinux.lds does not exist, and page.h looks
different.

However, I did change the 2 occurances of 0xc0000000 into 0x90000000:

#ifdef __ASSEMBLY__
#define __PAGE_OFFSET		(0x90000000)
#else
#define __PAGE_OFFSET		(0x90000000UL)
#endif

After recompiling the kernel, both qemu-fast and qemu refuses to boot
this kernel:

qemu -nographic -hda linux.img -kernel ../linux-qemu/arch/i386/boot/bzImage -append "root=/dev/hda ide2=noprobe ide3=noprobe ide4=noprobe ide5=noprobe"

...boots not bzImage, but the kernel which resides in the image file
linux.img.

ls -l ../linux-qemu/arch/i386/boot/bzImage  gives:
-rw-rw-r--    1 rutger   rutger     948860 Jan  5 21:37 ../linux-qemu/arch/i386/boot/bzImage

Do I need to change something different in 2.6?

-- 
Rutger Nijlunsing ----------------------------- rutger @ tux.tmfweb.nl
never attribute to a conspiracy which can be explained by incompetence
----------------------------------------------------------------------

Re: [Qemu-devel] qemu on 2.6.0-rc1

[J. Mayer]

On Mon, 2004-01-05 at 23:31, Rutger Nijlunsing wrote:
> Hi,
> 
> I try to run qemu-fast on Linux kernel 2.6.0-rc1. I could not find the
> things which I should change in the documentation:
....
> After recompiling the kernel, both qemu-fast and qemu refuses to boot
> this kernel:
> 
> qemu -nographic -hda linux.img -kernel ../linux-qemu/arch/i386/boot/bzImage -append "root=/dev/hda ide2=noprobe ide3=noprobe ide4=noprobe ide5=noprobe"
> 
> ...boots not bzImage, but the kernel which resides in the image file
> linux.img.

The fact is qemu -kernel relies on the fact that the kernel has a
integrated bootloader at the start of the bzImage. This has been
suppressed in 2.6 kernels, so you should install your kernel inside a
loop device install a MBR at the start of this device. You can use lilo
to do this.

-- 
J. Mayer <l_indien@magic.fr>
Never organized

Re: [Qemu-devel] qemu on 2.6.0-rc1

[Rusty Russell]

In message <20040105223155.GA14669@best.ms.philips.com> you write:
> Hi,
> 
> I try to run qemu-fast on Linux kernel 2.6.0-rc1. I could not find the
> things which I should change in the documentation:

Here's my patch, included below.  2.6.0 requires the vsyscall area to
move as well.

Sure, you can run the non-fast qemu, but it's MUCH slower.

Andrew?   This patch also has the benifit of moving all the constants
to one place, as well as adding a CONFIG_QEMU option.  I use it for
testing all the time.

Thanks,
Rusty.

Name: QEMU CONFIG Patch
Author: Rusty Russell
Status: Tested on 2.6.0

D: Convenient patch based on Fabrice Bellard's documentation on how to
D: boot an x86 kernel under qemu.  Adjusts PAGE_OFFSET, HZ, and
D: FIXADDR_TOP.

diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32324-linux-2.6.0/arch/i386/Kconfig .32324-linux-2.6.0.updated/arch/i386/Kconfig
--- .32324-linux-2.6.0/arch/i386/Kconfig	2003-10-09 18:02:48.000000000 +1000
+++ .32324-linux-2.6.0.updated/arch/i386/Kconfig	2003-12-26 16:46:49.000000000 +1100
@@ -307,6 +307,14 @@ config X86_GENERIC
 	  when it has moderate overhead. This is intended for generic 
 	  distributions kernels.
 
+config QEMU
+	bool "Kernel to run under QEMU"
+	depends on EXPERIMENTAL
+	help
+	  Select this if you want to boot the kernel inside qemu-fast,
+	  the non-mmu version of the x86 emulator.  See
+	  <http://fabrice.bellard.free.fr/qemu/>.  Say N.
+
 #
 # Define implied options from the CPU selection here
 #
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32324-linux-2.6.0/arch/i386/kernel/Makefile .32324-linux-2.6.0.updated/arch/i386/kernel/Makefile
--- .32324-linux-2.6.0/arch/i386/kernel/Makefile	2003-09-29 10:25:15.000000000 +1000
+++ .32324-linux-2.6.0.updated/arch/i386/kernel/Makefile	2003-12-26 16:46:49.000000000 +1100
@@ -46,12 +46,14 @@ quiet_cmd_syscall = SYSCALL $@
       cmd_syscall = $(CC) -nostdlib $(SYSCFLAGS_$(@F)) \
 		          -Wl,-T,$(filter-out FORCE,$^) -o $@
 
+export AFLAGS_vsyscall.lds.o += -P -C -U$(ARCH)
+
 vsyscall-flags = -shared -s -Wl,-soname=linux-gate.so.1
 SYSCFLAGS_vsyscall-sysenter.so	= $(vsyscall-flags)
 SYSCFLAGS_vsyscall-int80.so	= $(vsyscall-flags)
 
 $(obj)/vsyscall-int80.so $(obj)/vsyscall-sysenter.so: \
-$(obj)/vsyscall-%.so: $(src)/vsyscall.lds $(obj)/vsyscall-%.o FORCE
+$(obj)/vsyscall-%.so: $(src)/vsyscall.lds.s $(obj)/vsyscall-%.o FORCE
 	$(call if_changed,syscall)
 
 # We also create a special relocatable object that should mirror the symbol
@@ -62,5 +64,5 @@ $(obj)/built-in.o: $(obj)/vsyscall-syms.
 $(obj)/built-in.o: ld_flags += -R $(obj)/vsyscall-syms.o
 
 SYSCFLAGS_vsyscall-syms.o = -r
-$(obj)/vsyscall-syms.o: $(src)/vsyscall.lds $(obj)/vsyscall-sysenter.o FORCE
+$(obj)/vsyscall-syms.o: $(src)/vsyscall.lds.s $(obj)/vsyscall-sysenter.o FORCE
 	$(call if_changed,syscall)
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32324-linux-2.6.0/arch/i386/kernel/vmlinux.lds.S .32324-linux-2.6.0.updated/arch/i386/kernel/vmlinux.lds.S
--- .32324-linux-2.6.0/arch/i386/kernel/vmlinux.lds.S	2003-09-22 10:27:28.000000000 +1000
+++ .32324-linux-2.6.0.updated/arch/i386/kernel/vmlinux.lds.S	2003-12-26 16:46:49.000000000 +1100
@@ -3,6 +3,7 @@
  */
 
 #include <asm-generic/vmlinux.lds.h>
+#include <asm/page.h>
 	
 OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
 OUTPUT_ARCH(i386)
@@ -10,7 +11,7 @@ ENTRY(startup_32)
 jiffies = jiffies_64;
 SECTIONS
 {
-  . = 0xC0000000 + 0x100000;
+  . = __PAGE_OFFSET + 0x100000;
   /* read-only */
   _text = .;			/* Text and read-only data */
   .text : {
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32324-linux-2.6.0/arch/i386/kernel/vsyscall.lds .32324-linux-2.6.0.updated/arch/i386/kernel/vsyscall.lds
--- .32324-linux-2.6.0/arch/i386/kernel/vsyscall.lds	2003-09-22 10:07:26.000000000 +1000
+++ .32324-linux-2.6.0.updated/arch/i386/kernel/vsyscall.lds	1970-01-01 10:00:00.000000000 +1000
@@ -1,67 +0,0 @@
-/*
- * Linker script for vsyscall DSO.  The vsyscall page is an ELF shared
- * object prelinked to its virtual address, and with only one read-only
- * segment (that fits in one page).  This script controls its layout.
- */
-
-/* This must match <asm/fixmap.h>.  */
-VSYSCALL_BASE = 0xffffe000;
-
-SECTIONS
-{
-  . = VSYSCALL_BASE + SIZEOF_HEADERS;
-
-  .hash           : { *(.hash) }		:text
-  .dynsym         : { *(.dynsym) }
-  .dynstr         : { *(.dynstr) }
-  .gnu.version    : { *(.gnu.version) }
-  .gnu.version_d  : { *(.gnu.version_d) }
-  .gnu.version_r  : { *(.gnu.version_r) }
-
-  /* This linker script is used both with -r and with -shared.
-     For the layouts to match, we need to skip more than enough
-     space for the dynamic symbol table et al.  If this amount
-     is insufficient, ld -shared will barf.  Just increase it here.  */
-  . = VSYSCALL_BASE + 0x400;
-
-  .text           : { *(.text) }		:text =0x90909090
-
-  .eh_frame_hdr   : { *(.eh_frame_hdr) }	:text :eh_frame_hdr
-  .eh_frame       : { KEEP (*(.eh_frame)) }	:text
-  .dynamic        : { *(.dynamic) }		:text :dynamic
-  .useless        : {
-  	*(.got.plt) *(.got)
-	*(.data .data.* .gnu.linkonce.d.*)
-	*(.dynbss)
-	*(.bss .bss.* .gnu.linkonce.b.*)
-  }						:text
-}
-
-/*
- * We must supply the ELF program headers explicitly to get just one
- * PT_LOAD segment, and set the flags explicitly to make segments read-only.
- */
-PHDRS
-{
-  text PT_LOAD FILEHDR PHDRS FLAGS(5); /* PF_R|PF_X */
-  dynamic PT_DYNAMIC FLAGS(4); /* PF_R */
-  eh_frame_hdr 0x6474e550; /* PT_GNU_EH_FRAME, but ld doesn't match the name */
-}
-
-/*
- * This controls what symbols we export from the DSO.
- */
-VERSION
-{
-  LINUX_2.5 {
-    global:
-    	__kernel_vsyscall;
-    	__kernel_sigreturn;
-    	__kernel_rt_sigreturn;
-
-    local: *;
-  };
-}
-
-/* The ELF entry point can be used to set the AT_SYSINFO value.  */
-ENTRY(__kernel_vsyscall);
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32324-linux-2.6.0/arch/i386/kernel/vsyscall.lds.S .32324-linux-2.6.0.updated/arch/i386/kernel/vsyscall.lds.S
--- .32324-linux-2.6.0/arch/i386/kernel/vsyscall.lds.S	1970-01-01 10:00:00.000000000 +1000
+++ .32324-linux-2.6.0.updated/arch/i386/kernel/vsyscall.lds.S	2003-12-26 16:46:49.000000000 +1100
@@ -0,0 +1,67 @@
+/*
+ * Linker script for vsyscall DSO.  The vsyscall page is an ELF shared
+ * object prelinked to its virtual address, and with only one read-only
+ * segment (that fits in one page).  This script controls its layout.
+ */
+#include <asm/fixmap.h>
+	
+VSYSCALL_BASE = __FIXADDR_TOP - 0x1000;
+
+SECTIONS
+{
+  . = VSYSCALL_BASE + SIZEOF_HEADERS;
+
+  .hash           : { *(.hash) }		:text
+  .dynsym         : { *(.dynsym) }
+  .dynstr         : { *(.dynstr) }
+  .gnu.version    : { *(.gnu.version) }
+  .gnu.version_d  : { *(.gnu.version_d) }
+  .gnu.version_r  : { *(.gnu.version_r) }
+
+  /* This linker script is used both with -r and with -shared.
+     For the layouts to match, we need to skip more than enough
+     space for the dynamic symbol table et al.  If this amount
+     is insufficient, ld -shared will barf.  Just increase it here.  */
+  . = VSYSCALL_BASE + 0x400;
+
+  .text           : { *(.text) }		:text =0x90909090
+
+  .eh_frame_hdr   : { *(.eh_frame_hdr) }	:text :eh_frame_hdr
+  .eh_frame       : { KEEP (*(.eh_frame)) }	:text
+  .dynamic        : { *(.dynamic) }		:text :dynamic
+  .useless        : {
+  	*(.got.plt) *(.got)
+	*(.data .data.* .gnu.linkonce.d.*)
+	*(.dynbss)
+	*(.bss .bss.* .gnu.linkonce.b.*)
+  }						:text
+}
+
+/*
+ * We must supply the ELF program headers explicitly to get just one
+ * PT_LOAD segment, and set the flags explicitly to make segments read-only.
+ */
+PHDRS
+{
+  text PT_LOAD FILEHDR PHDRS FLAGS(5); /* PF_R|PF_X */
+  dynamic PT_DYNAMIC FLAGS(4); /* PF_R */
+  eh_frame_hdr 0x6474e550; /* PT_GNU_EH_FRAME, but ld doesn't match the name */
+}
+
+/*
+ * This controls what symbols we export from the DSO.
+ */
+VERSION
+{
+  LINUX_2.5 {
+    global:
+    	__kernel_vsyscall;
+    	__kernel_sigreturn;
+    	__kernel_rt_sigreturn;
+
+    local: *;
+  };
+}
+
+/* The ELF entry point can be used to set the AT_SYSINFO value.  */
+ENTRY(__kernel_vsyscall);
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32324-linux-2.6.0/include/asm-i386/fixmap.h .32324-linux-2.6.0.updated/include/asm-i386/fixmap.h
--- .32324-linux-2.6.0/include/asm-i386/fixmap.h	2003-09-22 10:09:12.000000000 +1000
+++ .32324-linux-2.6.0.updated/include/asm-i386/fixmap.h	2003-12-26 16:46:49.000000000 +1100
@@ -14,6 +14,19 @@
 #define _ASM_FIXMAP_H
 
 #include <linux/config.h>
+
+/* used by vmalloc.c, vsyscall.lds.S.
+ *
+ * Leave one empty page between vmalloc'ed areas and
+ * the start of the fixmap.
+ */
+#ifdef CONFIG_QEMU
+#define __FIXADDR_TOP	0xa7fff000
+#else
+#define __FIXADDR_TOP	0xfffff000
+#endif
+
+#ifndef __ASSEMBLY__
 #include <linux/kernel.h>
 #include <asm/acpi.h>
 #include <asm/apicdef.h>
@@ -94,13 +107,8 @@ extern void __set_fixmap (enum fixed_add
 #define clear_fixmap(idx) \
 		__set_fixmap(idx, 0, __pgprot(0))
 
-/*
- * used by vmalloc.c.
- *
- * Leave one empty page between vmalloc'ed areas and
- * the start of the fixmap.
- */
-#define FIXADDR_TOP	(0xfffff000UL)
+#define FIXADDR_TOP	((unsigned long)__FIXADDR_TOP)
+
 #define __FIXADDR_SIZE	(__end_of_permanent_fixed_addresses << PAGE_SHIFT)
 #define FIXADDR_START	(FIXADDR_TOP - __FIXADDR_SIZE)
 
@@ -145,4 +153,5 @@ static inline unsigned long virt_to_fix(
 	return __virt_to_fix(vaddr);
 }
 
+#endif /* !__ASSEMBLY__ */
 #endif
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32324-linux-2.6.0/include/asm-i386/page.h .32324-linux-2.6.0.updated/include/asm-i386/page.h
--- .32324-linux-2.6.0/include/asm-i386/page.h	2003-09-22 10:06:42.000000000 +1000
+++ .32324-linux-2.6.0.updated/include/asm-i386/page.h	2003-12-26 16:46:49.000000000 +1100
@@ -10,10 +10,10 @@
 #define LARGE_PAGE_SIZE (1UL << PMD_SHIFT)
 
 #ifdef __KERNEL__
-#ifndef __ASSEMBLY__
-
 #include <linux/config.h>
 
+#ifndef __ASSEMBLY__
+
 #ifdef CONFIG_X86_USE_3DNOW
 
 #include <asm/mmx.h>
@@ -115,12 +115,19 @@ static __inline__ int get_order(unsigned
 #endif /* __ASSEMBLY__ */
 
 #ifdef __ASSEMBLY__
+#ifdef CONFIG_QEMU
+#define __PAGE_OFFSET		(0x90000000)
+#else
 #define __PAGE_OFFSET		(0xC0000000)
+#endif /* QEMU */
+#else
+#ifdef CONFIG_QEMU
+#define __PAGE_OFFSET		(0x90000000UL)
 #else
 #define __PAGE_OFFSET		(0xC0000000UL)
+#endif /* QEMU */
 #endif
 
-
 #define PAGE_OFFSET		((unsigned long)__PAGE_OFFSET)
 #define VMALLOC_RESERVE		((unsigned long)__VMALLOC_RESERVE)
 #define MAXMEM			(-__PAGE_OFFSET-__VMALLOC_RESERVE)
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32324-linux-2.6.0/include/asm-i386/param.h .32324-linux-2.6.0.updated/include/asm-i386/param.h
--- .32324-linux-2.6.0/include/asm-i386/param.h	2003-09-21 17:26:06.000000000 +1000
+++ .32324-linux-2.6.0.updated/include/asm-i386/param.h	2003-12-26 16:46:49.000000000 +1100
@@ -2,7 +2,12 @@
 #define _ASMi386_PARAM_H
 
 #ifdef __KERNEL__
-# define HZ		1000		/* Internal kernel timer frequency */
+# include <linux/config.h>
+# ifdef CONFIG_QEMU
+#  define HZ		100
+# else
+#  define HZ		1000		/* Internal kernel timer frequency */
+# endif
 # define USER_HZ	100		/* .. some user interfaces are in "ticks" */
 # define CLOCKS_PER_SEC	(USER_HZ)	/* like times() */
 #endif
--
  Anyone who quotes me in their sig is an idiot. -- Rusty Russell.

Re: [Qemu-devel] qemu on 2.6.0-rc1

[Fabrice Bellard]

It seems I need to update the documentation :-)

Did you try to submit the non QEMU specific patches to Linus ? I think 
it is dangerous to leave harcoded constants in somes parts of the kernel.

BTW, in the current CVS, qemu-fast is about 2.3 times faster as the 
previous version (still 9 times slower than native on a P2 500 MHz on a 
typical 'gcc -pipe' compilation task which involves a lot of task 
switches).

I am planning simple x86 host specific optimisations to get closer to 
native performance... and SMP is getting closer now with my last changes.

Fabrice.

Rusty Russell wrote:
> In message <20040105223155.GA14669@best.ms.philips.com> you write:
> 
>> Hi,
>> 
>> I try to run qemu-fast on Linux kernel 2.6.0-rc1. I could not find
>> the things which I should change in the documentation:
> 
> 
> Here's my patch, included below.  2.6.0 requires the vsyscall area to
>  move as well.
> 
> Sure, you can run the non-fast qemu, but it's MUCH slower.
> 
> Andrew?   This patch also has the benifit of moving all the constants
>  to one place, as well as adding a CONFIG_QEMU option.  I use it for 
> testing all the time.
> 
> Thanks, Rusty.

Re: [Qemu-devel] qemu on 2.6.0-rc1

[Rusty Russell]

In message <3FFB2D46.90002@free.fr> you write:
> It seems I need to update the documentation :-)
> 
> Did you try to submit the non QEMU specific patches to Linus ? I think 
> it is dangerous to leave harcoded constants in somes parts of the kernel.

I'll see if Andrew takes the whole thing first.  If not, I'll submit
the "cleanup" part by itself.  I'd really like the whole thing.

> BTW, in the current CVS, qemu-fast is about 2.3 times faster as the 
> previous version (still 9 times slower than native on a P2 500 MHz on a 
> typical 'gcc -pipe' compilation task which involves a lot of task 
> switches).

0.5.1 feels about 10x slower booting the kernel (didn't measure, but
it's that order of mag).  Under 3 would really sway me...

> and SMP is getting closer now with my last changes.

Any idea when it'll be testable?  And do you prefer cash or check?

Rusty.
--
  Anyone who quotes me in their sig is an idiot. -- Rusty Russell.

Re: [Qemu-devel] qemu on 2.6.0-rc1

[Herbert Poetzl]

On Wed, Jan 07, 2004 at 12:53:40PM +1100, Rusty Russell wrote:
> In message <3FFB2D46.90002@free.fr> you write:
> > It seems I need to update the documentation :-)
> > 
> > Did you try to submit the non QEMU specific patches to Linus ? I think 
> > it is dangerous to leave harcoded constants in somes parts of the kernel.
> 
> I'll see if Andrew takes the whole thing first.  If not, I'll submit
> the "cleanup" part by itself.  I'd really like the whole thing.
> 
> > BTW, in the current CVS, qemu-fast is about 2.3 times faster as the 
> > previous version (still 9 times slower than native on a P2 500 MHz on a 
> > typical 'gcc -pipe' compilation task which involves a lot of task 
> > switches).
> 
> 0.5.1 feels about 10x slower booting the kernel (didn't measure, but
> it's that order of mag).  Under 3 would really sway me...
> 
> > and SMP is getting closer now with my last changes.
> 
> Any idea when it'll be testable?  And do you prefer cash or check?

I'm willing to do extensive testing on SMP capable
qemu, maybe ever gdb support?

I had some nightmares trying to get bochs up and
running in SMP emulation mode, and finally gave
up (too many apic/tsc issues) ...

TIA,
Herbert

> Rusty.
> --
>   Anyone who quotes me in their sig is an idiot. -- Rusty Russell.
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://mail.nongnu.org/mailman/listinfo/qemu-devel

Re: [Qemu-devel] qemu on 2.6.0-rc1

[Fabrice Bellard]

Herbert Poetzl wrote:

> I'm willing to do extensive testing on SMP capable
> qemu, maybe ever gdb support?

In the current CVS version gdb is less usable as it allows you to 
inspect  physical memory only. Is there a standard way in gdb to switch 
between physical and virtual memory dumps ?

Fabrice.

Re: [Qemu-devel] qemu on 2.6.0-rc1

[Fabrice Bellard]

Rusty Russell wrote:

>>and SMP is getting closer now with my last changes.
> 
> 
> Any idea when it'll be testable?  And do you prefer cash or check?

If I was paid it would go faster :-) All I can say now is that it won't 
come before I can manage to install and boot Windows XP in softmmu mode.

Fabrice.

Re: [Qemu-devel] qemu on 2.6.0-rc1

[Daniel Jacobowitz]

On Wed, Jan 07, 2004 at 10:33:57PM +0100, Fabrice Bellard wrote:
> Herbert Poetzl wrote:
> 
> >I'm willing to do extensive testing on SMP capable
> >qemu, maybe ever gdb support?
> 
> In the current CVS version gdb is less usable as it allows you to 
> inspect  physical memory only. Is there a standard way in gdb to switch 
> between physical and virtual memory dumps ?

Nope.  GDB does not know anything about physical or virtual memory; it
displays whatever the remote stub happens to provide.

I'd think virtual would be more useful...

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer