similar to: [PATCH 11/21] i386 Stop deleting nt

Fairly straight code motion. Split non-virtualizable pieces of processor.h into default and VMI subarches. CPUID is non-virtualizable, since it doesn't trap, and very often the hypervisor will want to hide specific feature bits from the kernel. To provide a replacement for call sites that use CPUID as a serializing instruction, the sync_core() macro is still available. Signed-off-by:

Fairly straight code motion. Split non-virtualizable pieces of processor.h into default and VMI subarches. CPUID is non-virtualizable, since it doesn't trap, and very often the hypervisor will want to hide specific feature bits from the kernel. To provide a replacement for call sites that use CPUID as a serializing instruction, the sync_core() macro is still available. Signed-off-by:

This version over last version: (1) Gets rid of the no_paravirt.h header and leaves native ops in place (with some reshuffling to keep then under one #ifdef). (2) Fixes the "X crashes with CONFIG_PARAVIRT=y" bug. (3) Puts __ex_table entry in paravirt iret. Another followup patch implements binary patching... Rusty. === Create a paravirt.h header for all the critical operations which

This version over last version: (1) Gets rid of the no_paravirt.h header and leaves native ops in place (with some reshuffling to keep then under one #ifdef). (2) Fixes the "X crashes with CONFIG_PARAVIRT=y" bug. (3) Puts __ex_table entry in paravirt iret. Another followup patch implements binary patching... Rusty. === Create a paravirt.h header for all the critical operations which

plain text document attachment (xx-paravirt-boot.patch) Boot up code modifications to get paravirt ops running. Signed-off-by: Steven Rostedt srostedt@redhat.com Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Index: clean-start/arch/x86_64/kernel/head.S =================================================================== --- clean-start.orig/arch/x86_64/kernel/head.S +++

plain text document attachment (xx-paravirt-boot.patch) Boot up code modifications to get paravirt ops running. Signed-off-by: Steven Rostedt srostedt@redhat.com Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Index: clean-start/arch/x86_64/kernel/head.S =================================================================== --- clean-start.orig/arch/x86_64/kernel/head.S +++

(Andrew, please sit these in the -mm tree for cooking) Create a paravirt.h header for (almost) all the critical operations which need to be replaced with hypervisor calls. For the moment, this simply includes no_paravirt.h, where all the native implementations now live. Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Zachary Amsden <zach@vmware.com>

(Andrew, please sit these in the -mm tree for cooking) Create a paravirt.h header for (almost) all the critical operations which need to be replaced with hypervisor calls. For the moment, this simply includes no_paravirt.h, where all the native implementations now live. Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Zachary Amsden <zach@vmware.com>

OK, at least two patches got dropped on the way from the mm tree to Andi's tree: the desc.h cleanup, and the processor.h rearrangement. Merging into Andi's tree without these patches must have been a nightmare 8( Andi then tried to fix it with x86_64-mm-paravirt-compile.patch but then it didn't boot so he disabled it in x86_64-mm-paravirt-broken.patch This patch undoes those two

OK, at least two patches got dropped on the way from the mm tree to Andi's tree: the desc.h cleanup, and the processor.h rearrangement. Merging into Andi's tree without these patches must have been a nightmare 8( Andi then tried to fix it with x86_64-mm-paravirt-compile.patch but then it didn't boot so he disabled it in x86_64-mm-paravirt-broken.patch This patch undoes those two

Fairly straightforward code motion in system.h into the sub-arch layer. Affected functionality include control register accessors, which are virtualizable but with great overhead due to the #GP cost; wbinvd, and most importantly, halt and interrupt control, which is non-virtualizable. Since read_cr4_safe can never fault on a VMI kernel (P5+ processor is required for VMI), we can omit the fault

Fairly straightforward code motion in system.h into the sub-arch layer. Affected functionality include control register accessors, which are virtualizable but with great overhead due to the #GP cost; wbinvd, and most importantly, halt and interrupt control, which is non-virtualizable. Since read_cr4_safe can never fault on a VMI kernel (P5+ processor is required for VMI), we can omit the fault

I discovered an even more subtle problem; the PnP BIOS code is saving the %fs and %gs segments in inline assembler, yet it also uses the same hack for patching in a fake real mode selector for the BIOS data area. Note that the protected mode selector 0x40 overlaps the user TLS area in the GDT; this means that badly timed PnP BIOS calls could come in, save %fs, come back, and restore %fs -- to

I discovered an even more subtle problem; the PnP BIOS code is saving the %fs and %gs segments in inline assembler, yet it also uses the same hack for patching in a fake real mode selector for the BIOS data area. Note that the protected mode selector 0x40 overlaps the user TLS area in the GDT; this means that badly timed PnP BIOS calls could come in, save %fs, come back, and restore %fs -- to

The per-CPU initialization code is copying in bogus data into thread->tls_array. Note that it copies &per_cpu(cpu_gdt_table, cpu), not &per_cpu(cpu_gdt_table, cpu)[GDT_ENTRY_TLS_MIN). That is totally broken and unnecessary. Make the initialization explicitly NULL. Signed-off-by: Zachary Amsden <zach@vmware.com> Index: linux-2.6.14-rc1/arch/i386/kernel/cpu/common.c

The per-CPU initialization code is copying in bogus data into thread->tls_array. Note that it copies &per_cpu(cpu_gdt_table, cpu), not &per_cpu(cpu_gdt_table, cpu)[GDT_ENTRY_TLS_MIN). That is totally broken and unnecessary. Make the initialization explicitly NULL. Signed-off-by: Zachary Amsden <zach@vmware.com> Index: linux-2.6.14-rc1/arch/i386/kernel/cpu/common.c

Add an accessor function for getting the per-CPU gdt. Callee must already have the CPU. Signed-off-by: Zachary Amsden <zach@vmware.com> Index: linux-2.6.14-rc1/include/asm-i386/desc.h =================================================================== --- linux-2.6.14-rc1.orig/include/asm-i386/desc.h 2005-09-20 14:49:10.000000000 -0700 +++ linux-2.6.14-rc1/include/asm-i386/desc.h

Add an accessor function for getting the per-CPU gdt. Callee must already have the CPU. Signed-off-by: Zachary Amsden <zach@vmware.com> Index: linux-2.6.14-rc1/include/asm-i386/desc.h =================================================================== --- linux-2.6.14-rc1.orig/include/asm-i386/desc.h 2005-09-20 14:49:10.000000000 -0700 +++ linux-2.6.14-rc1/include/asm-i386/desc.h

Ancient inline assembler that manipulates descriptor tables is unreadable and has no type checking. Doing this in C actually generates better code, saves code space, and improves readability. The fact that you must cast descriptors to (char *) for the inline assembler to work properly caused me no end of grief working on these patches. Note that GCC does not generate rotations to utilize

Ancient inline assembler that manipulates descriptor tables is unreadable and has no type checking. Doing this in C actually generates better code, saves code space, and improves readability. The fact that you must cast descriptors to (char *) for the inline assembler to work properly caused me no end of grief working on these patches. Note that GCC does not generate rotations to utilize