On Fri, Oct 14, 2016 at 02:05:14PM -0400, Boris Ostrovsky wrote:
> Start PVH guest at XEN_ELFNOTE_PHYS32_ENTRY address. Setup hypercall
> page, initialize boot_params, enable early page tables.
> 
> Since this stub is executed before kernel entry point we cannot use
> variables in .bss which is cleared by kernel. We explicitly place
> variables that are initialized here into .data.
> 
> Signed-off-by: Boris Ostrovsky <boris.ostrov...@oracle.com>
> Signed-off-by: Matt Fleming <m...@codeblueprint.co.uk>
> ---
>  arch/x86/xen/Kconfig     |   2 +-
>  arch/x86/xen/Makefile    |   1 +
>  arch/x86/xen/enlighten.c |  87 +++++++++++++++++++++++++++-
>  arch/x86/xen/xen-pvh.S   | 143 
> +++++++++++++++++++++++++++++++++++++++++++++++
>  include/xen/xen.h        |   5 ++
>  5 files changed, 236 insertions(+), 2 deletions(-)
>  create mode 100644 arch/x86/xen/xen-pvh.S
> 
> diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig
> index c7b15f3..76b6dbd 100644
> --- a/arch/x86/xen/Kconfig
> +++ b/arch/x86/xen/Kconfig
> @@ -53,5 +53,5 @@ config XEN_DEBUG_FS
>  
>  config XEN_PVH
>       bool "Support for running as a PVH guest"
> -     depends on X86_64 && XEN && XEN_PVHVM
> +     depends on XEN && XEN_PVHVM && ACPI
>       def_bool n
> diff --git a/arch/x86/xen/Makefile b/arch/x86/xen/Makefile
> index e47e527..cb0164a 100644
> --- a/arch/x86/xen/Makefile
> +++ b/arch/x86/xen/Makefile
> @@ -23,3 +23,4 @@ obj-$(CONFIG_XEN_DEBUG_FS)  += debugfs.o
>  obj-$(CONFIG_XEN_DOM0)               += vga.o
>  obj-$(CONFIG_SWIOTLB_XEN)    += pci-swiotlb-xen.o
>  obj-$(CONFIG_XEN_EFI)                += efi.o
> +obj-$(CONFIG_XEN_PVH)                += xen-pvh.o
> diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
> index dc4ed0c..d38d568 100644
> --- a/arch/x86/xen/enlighten.c
> +++ b/arch/x86/xen/enlighten.c
> @@ -45,6 +45,7 @@
>  #include <xen/interface/memory.h>
>  #include <xen/interface/nmi.h>
>  #include <xen/interface/xen-mca.h>
> +#include <xen/interface/hvm/start_info.h>
>  #include <xen/features.h>
>  #include <xen/page.h>
>  #include <xen/hvm.h>
> @@ -121,7 +122,8 @@
>  DEFINE_PER_CPU(uint32_t, xen_vcpu_id);
>  EXPORT_PER_CPU_SYMBOL(xen_vcpu_id);
>  
> -enum xen_domain_type xen_domain_type = XEN_NATIVE;
> +enum xen_domain_type xen_domain_type
> +     __attribute__((section(".data"))) = XEN_NATIVE;
>  EXPORT_SYMBOL_GPL(xen_domain_type);
>  
>  unsigned long *machine_to_phys_mapping = (void *)MACH2PHYS_VIRT_START;
> @@ -176,6 +178,17 @@ struct tls_descs {
>   */
>  static DEFINE_PER_CPU(struct tls_descs, shadow_tls_desc);
>  
> +#ifdef CONFIG_XEN_PVH
> +/*
> + * PVH variables. These need to live in data segment since they are
> + * initialized before startup_{32|64}, which clear .bss, are invoked.
> + */
> +int xen_pvh __attribute__((section(".data"))) = 0;

unsigned int?
> +struct hvm_start_info pvh_start_info __attribute__((section(".data")));
> +uint pvh_start_info_sz = sizeof(pvh_start_info);

unsigned int please. Typedefs in Linux are frowned upon.

> +struct boot_params pvh_bootparams __attribute__((section(".data")));
> +#endif
> +
>  static void clamp_max_cpus(void)
>  {
>  #ifdef CONFIG_SMP
> @@ -1669,6 +1682,78 @@ asmlinkage __visible void __init xen_start_kernel(void)
>  #endif
>  }
>  
> +#ifdef CONFIG_XEN_PVH
> +static void __init init_pvh_bootparams(void)
> +{
> +     struct xen_memory_map memmap;
> +     int i;

unsigned int?
> +
> +     memset(&pvh_bootparams, 0, sizeof(pvh_bootparams));
> +
> +     memmap.nr_entries = ARRAY_SIZE(pvh_bootparams.e820_map);
> +     set_xen_guest_handle(memmap.buffer, pvh_bootparams.e820_map);
> +     if (HYPERVISOR_memory_op(XENMEM_memory_map, &memmap)) {
> +             xen_raw_console_write("XENMEM_memory_map failed\n");

Should we print the error value at least?
> +             BUG();
> +     }
> +
> +     pvh_bootparams.e820_map[memmap.nr_entries].addr =
> +             ISA_START_ADDRESS;

What if nr_entries is 128? Should we double-check for that?

> +     pvh_bootparams.e820_map[memmap.nr_entries].size =
> +             ISA_END_ADDRESS - ISA_START_ADDRESS;
> +     pvh_bootparams.e820_map[memmap.nr_entries++].type =
> +             E820_RESERVED;
> +
> +     sanitize_e820_map(pvh_bootparams.e820_map,
> +                       ARRAY_SIZE(pvh_bootparams.e820_map),
> +                       &memmap.nr_entries); 
> +
> +     pvh_bootparams.e820_entries = memmap.nr_entries;
> +     for (i = 0; i < pvh_bootparams.e820_entries; i++)
> +             e820_add_region(pvh_bootparams.e820_map[i].addr,
> +                             pvh_bootparams.e820_map[i].size,
> +                             pvh_bootparams.e820_map[i].type);
> +
> +     pvh_bootparams.hdr.cmd_line_ptr =
> +             pvh_start_info.cmdline_paddr;
> +
> +     /* The first module is always ramdisk */

Could you add an period at end please?
> +     if (pvh_start_info.nr_modules) {
> +             struct hvm_modlist_entry *modaddr =
> +                     __va(pvh_start_info.modlist_paddr);
> +             pvh_bootparams.hdr.ramdisk_image = modaddr->paddr;
> +             pvh_bootparams.hdr.ramdisk_size = modaddr->size;
> +     }
> +
> +     /*
> +      * See Documentation/x86/boot.txt.
> +      *
> +      * Version 2.12 supports Xen entry point but we will use default x86/PC
> +      * environment (i.e. hardware_subarch 0).
> +      */
> +     pvh_bootparams.hdr.version = 0x212;
> +     pvh_bootparams.hdr.type_of_loader = (9 << 4) | 0; /* Xen loader */
> +}
> +
> +/*
> + * This routine (and those that it might call) should not use
> + * anything that lives in .bss since that segment will be cleared later

And maybe one here too?
> + */
> +void __init xen_prepare_pvh(void)
> +{
> +     u32 eax, ecx, edx, msr;

msr = 0 ?
> +     u64 pfn;
> +
> +     xen_pvh = 1;
> +
> +     cpuid(xen_cpuid_base() + 2, &eax, &msr, &ecx, &edx);

cpuid_ebx ? And that way you don't have have ecx and edx?
> +     pfn = __pa(hypercall_page);
> +     wrmsr_safe(msr, (u32)pfn, (u32)(pfn >> 32));
> +
> +     init_pvh_bootparams();
> +}
> +#endif
> +
>  void __ref xen_hvm_init_shared_info(void)
>  {
>       int cpu;
> diff --git a/arch/x86/xen/xen-pvh.S b/arch/x86/xen/xen-pvh.S
> new file mode 100644
> index 0000000..58c477b
> --- /dev/null
> +++ b/arch/x86/xen/xen-pvh.S
> @@ -0,0 +1,143 @@
> +/*
> + * Copyright C 2016, Oracle and/or its affiliates. All rights reserved.
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; either version 2 of the License, or
> + * (at your option) any later version.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + *
> + * You should have received a copy of the GNU General Public License along
> + * with this program.  If not, see <http://www.gnu.org/licenses/>.
> + */
> +
> +     .code32
> +     .text
> +#define _pa(x)          ((x) - __START_KERNEL_map)
> +
> +#include <linux/elfnote.h>
> +#include <linux/init.h>
> +#include <linux/linkage.h>
> +#include <asm/segment.h>
> +#include <asm/asm.h>
> +#include <asm/boot.h>
> +#include <asm/processor-flags.h>
> +#include <asm/msr.h>
> +#include <xen/interface/elfnote.h>
> +
> +     __HEAD
> +     .code32
> +
> +/* Entry point for PVH guests */
> +ENTRY(pvh_start_xen)

You are missing the ENDPROC macro at the end.

> +     cli
> +     cld
> +
> +     mov $_pa(gdt), %eax
> +     lgdt (%eax)
> +
> +     movl $(__BOOT_DS),%eax
> +     movl %eax,%ds
> +     movl %eax,%es
> +     movl %eax,%ss
> +
> +     /* Stash hvm_start_info */
> +     mov $_pa(pvh_start_info), %edi
> +     mov %ebx, %esi

Should we derference the first byte or such to check for the magic
string? Actually I am not even seeing the check in the C code?

> +     mov $_pa(pvh_start_info_sz), %ecx
> +     mov (%ecx), %ecx
> +     rep
> +     movsb
> +
> +     movl $_pa(early_stack_end), %eax
> +     movl %eax, %esp
> +
> +     /* Enable PAE mode */

Periods are nice! Truly!

> +     movl %cr4, %eax
> +     orl $X86_CR4_PAE, %eax
> +     movl %eax, %cr4
> +
> +#ifdef CONFIG_X86_64
> +     /* Enable Long mode */

:-) I think you know what I am going to say here.

> +     movl $MSR_EFER, %ecx
> +     rdmsr
> +     btsl $_EFER_LME, %eax
> +     wrmsr
> +
> +     /* Enable pre-constructed page tables */

And here.

> +     mov $_pa(init_level4_pgt), %eax
> +     movl %eax, %cr3
> +     movl $(X86_CR0_PG | X86_CR0_PE), %eax
> +     movl %eax, %cr0
> +
> +     /* Jump to 64-bit mode. */
> +     pushl $__KERNEL_CS
> +     leal _pa(1f), %eax
> +     pushl %eax
> +     lret
> +
> +     /* 64-bit entry point */

And right here.
> +     .code64
> +1:
> +     call xen_prepare_pvh
> +
> +     /* startup_64 expects boot_params in %rsi */

..
> +     mov $_pa(pvh_bootparams), %rsi
> +     movq $_pa(startup_64), %rax
> +     jmp *%rax
> +
> +#else /* CONFIG_X86_64 */
> +
> +     call setup_pgtable_32
> +
> +     mov $_pa(initial_page_table), %eax
> +     movl %eax, %cr3
> +
> +     movl %cr0, %eax
> +     orl $(X86_CR0_PG | X86_CR0_PE), %eax
> +     movl %eax, %cr0
> +
> +     ljmp $__BOOT_CS,$1f
> +1:
> +     call xen_prepare_pvh
> +     mov $_pa(pvh_bootparams), %esi
> +
> +     /* startup_32 doesn't expect paging and PAE to be on */

Should 'startup_32' be documented with this?


> +     ljmp $__BOOT_CS,$_pa(2f)
> +2:
> +     movl %cr0, %eax
> +     andl $~X86_CR0_PG, %eax
> +     movl %eax, %cr0
> +     movl %cr4, %eax
> +     andl $~X86_CR4_PAE, %eax
> +     movl %eax, %cr4
> +
> +     ljmp    $0x10, $_pa(startup_32)
> +#endif
> +
> +     .data
> +gdt:
> +     .word   gdt_end - gdt
> +     .long   _pa(gdt)
> +     .word   0
> +     .quad   0x0000000000000000 /* NULL descriptor */
> +#ifdef CONFIG_X86_64
> +     .quad   0x00af9a000000ffff /* __KERNEL_CS */
> +#else
> +     .quad   0x00cf9a000000ffff /* __KERNEL_CS */
> +#endif
> +     .quad   0x00cf92000000ffff /* __KERNEL_DS */
> +gdt_end:
> +
> +     .bss
> +     .balign 4
> +early_stack:
> +     .fill 16, 1, 0
> +early_stack_end:
> +
> +     ELFNOTE(Xen, XEN_ELFNOTE_PHYS32_ENTRY,
> +                  _ASM_PTR (pvh_start_xen - __START_KERNEL_map))
> diff --git a/include/xen/xen.h b/include/xen/xen.h
> index d0f9684..ed3f841 100644
> --- a/include/xen/xen.h
> +++ b/include/xen/xen.h
> @@ -29,6 +29,11 @@ enum xen_domain_type {
>  #define xen_initial_domain() (0)
>  #endif       /* CONFIG_XEN_DOM0 */
>  
> +#ifdef CONFIG_XEN_PVH
> +extern int xen_pvh;
> +#define xen_pvh_domain()     (xen_hvm_domain() && xen_pvh)
> +#else
>  #define xen_pvh_domain()     (0)
> +#endif
>  
>  #endif       /* _XEN_XEN_H */
> -- 
> 1.8.3.1
> 
> 
> _______________________________________________
> Xen-devel mailing list
> xen-de...@lists.xen.org
> https://lists.xen.org/xen-devel

Reply via email to