[RFC 4/5] ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching
Nicolas Pitre
nico at fluxnic.net
Tue Jan 4 16:27:37 EST 2011
On Tue, 4 Jan 2011, Russell King - ARM Linux wrote:
> This idea came from Nicolas, Eric Miao produced an initial version,
> which was then rewritten into this.
>
> Patch the physical to virtual translations at runtime. As we modify
> the code, this makes it incompatible with XIP kernels, but on allows
> is to achieve this with minimal loss of performance.
Above sentence looks weird.
> As many translations are of the form:
>
> physical = virtual + (PHYS_OFFSET - PAGE_OFFSET)
> virtual = physical - (PHYS_OFFSET - PAGE_OFFSET)
>
> we generate an 'add' instruction for __virt_to_phys(), and a 'sub'
> instruction for __phys_to_virt(). We calculate at run time (PHYS_OFFSET
> - PAGE_OFFSET) by comparing the address prior to MMU initialization with
> where it should be once the MMU has been initialized, and place this
> constant into the above add/sub instructions.
>
> Once we have (PHYS_OFFSET - PAGE_OFFSET), we can calcuate the real
s/calcuate/calculate/
> PHYS_OFFSET as PAGE_OFFSET is a build-time constant, and save this for
> the C-mode PHYS_OFFSET variable definition to use.
>
> At present, we are unable to support Realview with Sparsemem enabled
> as this uses a complex mapping function, and MSM as this requires a
> constant which will not fit in our math instruction.
>
> Signed-off-by: Russell King <rmk+kernel at arm.linux.org.uk>
[...]
> @@ -608,6 +621,7 @@ config ARCH_PXA
> select TICK_ONESHOT
> select PLAT_PXA
> select SPARSE_IRQ
> + select ARM_PATCH_PHYS_VIRT
> help
> Support for Intel/Marvell's PXA2xx/PXA3xx processor line.
>
This hunk should go away.
> diff --git a/arch/arm/include/asm/memory.h b/arch/arm/include/asm/memory.h
> index 00b04ae..288b690 100644
> --- a/arch/arm/include/asm/memory.h
> +++ b/arch/arm/include/asm/memory.h
> @@ -24,8 +24,6 @@
> */
> #define UL(x) _AC(x, UL)
>
> -#define PHYS_OFFSET PLAT_PHYS_OFFSET
> -
> #ifdef CONFIG_MMU
>
> /*
> @@ -135,16 +133,6 @@
> #endif
>
> /*
> - * Physical vs virtual RAM address space conversion. These are
> - * private definitions which should NOT be used outside memory.h
> - * files. Use virt_to_phys/phys_to_virt/__pa/__va instead.
> - */
> -#ifndef __virt_to_phys
> -#define __virt_to_phys(x) ((x) - PAGE_OFFSET + PHYS_OFFSET)
> -#define __phys_to_virt(x) ((x) - PHYS_OFFSET + PAGE_OFFSET)
> -#endif
> -
> -/*
> * Convert a physical address to a Page Frame Number and back
> */
> #define __phys_to_pfn(paddr) ((paddr) >> PAGE_SHIFT)
> @@ -159,6 +147,49 @@
> #ifndef __ASSEMBLY__
>
> /*
> + * Physical vs virtual RAM address space conversion. These are
> + * private definitions which should NOT be used outside memory.h
> + * files. Use virt_to_phys/phys_to_virt/__pa/__va instead.
> + */
> +#ifndef __virt_to_phys
> +#ifdef CONFIG_ARM_PATCH_PHYS_VIRT
> +
> +extern unsigned long __pv_phys_offset;
> +#define PHYS_OFFSET __pv_phys_offset
> +
> +#define __pv_stub(from,to,instr) \
> + __asm__("@ __pv_stub\n" \
> + "1: " instr " %0, %1, %2\n" \
> + " .pushsection .pv_table,\"a\"\n" \
> + " .long 1b\n" \
> + " .popsection\n" \
> + : "=r" (to) \
> + : "r" (from), "I" (1))
> +
> +static inline unsigned long __virt_to_phys(unsigned long x)
> +{
> + unsigned long t;
> + __pv_stub(x, t, "add");
> + return t;
> +}
> +
> +static inline unsigned long __phys_to_virt(unsigned long x)
> +{
> + unsigned long t;
> + __pv_stub(x, t, "sub");
> + return t;
> +}
> +#else
> +#define __virt_to_phys(x) ((x) - PAGE_OFFSET + PHYS_OFFSET)
> +#define __phys_to_virt(x) ((x) - PHYS_OFFSET + PAGE_OFFSET)
> +#endif
> +#endif
> +
> +#ifndef PHYS_OFFSET
> +#define PHYS_OFFSET PLAT_PHYS_OFFSET
> +#endif
> +
> +/*
> * The DMA mask corresponding to the maximum bus address allocatable
> * using GFP_DMA. The default here places no restriction on DMA
> * allocations. This must be the smallest DMA mask in the system,
> diff --git a/arch/arm/kernel/head.S b/arch/arm/kernel/head.S
> index aedd80e..258b0ca 100644
> --- a/arch/arm/kernel/head.S
> +++ b/arch/arm/kernel/head.S
> @@ -100,6 +100,9 @@ ENTRY(stext)
> #ifdef CONFIG_SMP_ON_UP
> bl __fixup_smp
> #endif
> +#ifdef CONFIG_ARM_PATCH_PHYS_VIRT
> + bl __fixup_pv_table
> +#endif
> bl __create_page_tables
>
> /*
> @@ -445,4 +448,64 @@ smp_on_up:
>
> #endif
>
> +#ifdef CONFIG_ARM_PATCH_PHYS_VIRT
> +
> +/* __fixup_pv_table - patch the stub instructions with the delta between
> + * PHYS_OFFSET and PAGE_OFFSET, which is assumed to be 16MiB aligned and
> + * can be expressed by an immediate shifter operand. The stub instruction
> + * has a form of '(add|sub) rd, rn, #imm'.
> + */
> +__fixup_pv_table:
> + adr r0, 1f
> + ldmia r0, {r3-r5,r7,ip}
> + sub r3, r0, r3 @ PHYS_OFFSET - PAGE_OFFSET
> + add r4, r4, r3 @ adjust table start address
> + add r5, r5, r3 @ adjust table end address
> + add ip, ip, r3 @ our PHYS_OFFSET
> + str ip, [r7, r3]! @ save to __pv_phys_offset
> + mov r6, r3, lsr #24 @ constant for add/sub instructions
> + teq r3, r6, lsl #24 @ must be 16MiB aligned
> + bne __error
> + orr r6, r6, #0x400 @ mask in rotate right 8 bits
> + str r6, [r7, #4] @ save to __pv_offset
Looks to me that the code up to now, including the referenced data
constants, can live in __init section. But the following code should be
in the .text section.
> +__fixup_a_pv_table:
> +2: cmp r4, r5
> + ldrlo r7, [r4], #4
> + ldrlo ip, [r7, r3]
> + bic ip, ip, #0x000000ff
> + bic ip, ip, #0x00000f00
> + orr ip, ip, r6
> + strlo ip, [r7, r3]
> + blo 2b
> + mov pc, lr
> +ENDPROC(__fixup_phys_virt)
> +
> +ENTRY(fixup_pv_table)
> + stmfd sp!, {r4 - r7, lr}
> + ldr r2, 2f @ get address of __pv_phys_offset
> + mov r3, #0 @ no offset
> + mov r4, r0 @ r0 = table start
> + add r5, r0, r1 @ r1 = table size
> + ldr r6, [r2, #4] @ get __pv_offset
> + bl __fixup_a_pv_table
> + ldmfd sp!, {r4 - r7, pc}
> +ENDPROC(fixup_pv_table)
> +
> + .align
> +1: .long .
> + .long __pv_table_begin
> + .long __pv_table_end
> +2: .long __pv_phys_offset
> + .long PAGE_OFFSET
> +
> + .data
> + .globl __pv_phys_offset
> + .type __pv_phys_offset, %object
> +__pv_phys_offset:
> + .long 0
> + .size __pv_phys_offset, . - __pv_phys_offset
> +__pv_offset:
> + .long 0
> +#endif
> +
> #include "head-common.S"
> diff --git a/arch/arm/kernel/module.c b/arch/arm/kernel/module.c
> index 0c1bb68..6a42e17 100644
> --- a/arch/arm/kernel/module.c
> +++ b/arch/arm/kernel/module.c
> @@ -276,12 +276,28 @@ struct mod_unwind_map {
> const Elf_Shdr *txt_sec;
> };
>
> +static const Elf_Shdr *find_mod_section(const Elf32_Ehdr *hdr,
> + const Elf_Shdr *sechdrs, const char *name)
> +{
> + const Elf_Shdr *s, *se;
> + const char *secstrs = (void *)hdr + sechdrs[hdr->e_shstrndx].sh_offset;
> +
> + for (s = sechdrs, se = sechdrs + hdr->e_shnum; s < se; s++)
> + if (strcmp(name, secstrs + s->sh_name) == 0)
> + return s;
> +
> + return NULL;
> +}
> +
> +extern void fixup_pv_table(const void *, unsigned long);
> +
> int module_finalize(const Elf32_Ehdr *hdr, const Elf_Shdr *sechdrs,
> struct module *mod)
> {
> + const Elf_Shdr *s = NULL;
> #ifdef CONFIG_ARM_UNWIND
> const char *secstrs = (void *)hdr + sechdrs[hdr->e_shstrndx].sh_offset;
> - const Elf_Shdr *s, *sechdrs_end = sechdrs + hdr->e_shnum;
> + const Elf_Shdr *sechdrs_end = sechdrs + hdr->e_shnum;
> struct mod_unwind_map maps[ARM_SEC_MAX];
> int i;
>
> @@ -323,6 +339,11 @@ int module_finalize(const Elf32_Ehdr *hdr, const Elf_Shdr *sechdrs,
> maps[i].txt_sec->sh_addr,
> maps[i].txt_sec->sh_size);
> #endif
> +#ifdef CONFIG_ARM_PATCH_PHYS_VIRT
> + s = find_mod_section(hdr, sechdrs, ".pv_table");
> + if (s)
> + fixup_pv_table((void *)s->sh_addr, s->sh_size);
> +#endif
> return 0;
> }
>
> diff --git a/arch/arm/kernel/setup.c b/arch/arm/kernel/setup.c
> index f67e682..7c5499d 100644
> --- a/arch/arm/kernel/setup.c
> +++ b/arch/arm/kernel/setup.c
> @@ -804,7 +804,7 @@ void __init setup_arch(char **cmdline_p)
> struct machine_desc *mdesc;
> char *from = default_command_line;
>
> - tags->mem.start = PHYS_OFFSET;
> + init_tags.mem.start = PHYS_OFFSET;
Doesn't this belong in 2/5 instead?
Nicolas
More information about the linux-arm-kernel
mailing list