linux_dsm_epyc7002/arch/arm/kernel/head.S

735 lines
19 KiB
ArmAsm
Raw Normal View History

/*
* linux/arch/arm/kernel/head.S
*
* Copyright (C) 1994-2002 Russell King
* Copyright (c) 2003 ARM Limited
* 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 version 2 as
* published by the Free Software Foundation.
*
* Kernel startup code for all 32-bit CPUs
*/
#include <linux/linkage.h>
#include <linux/init.h>
#include <asm/assembler.h>
#include <asm/cp15.h>
#include <asm/domain.h>
#include <asm/ptrace.h>
#include <asm/asm-offsets.h>
#include <asm/memory.h>
#include <asm/thread_info.h>
#include <asm/pgtable.h>
#if defined(CONFIG_DEBUG_LL) && !defined(CONFIG_DEBUG_SEMIHOSTING)
#include CONFIG_DEBUG_LL_INCLUDE
#endif
/*
* swapper_pg_dir is the virtual address of the initial page table.
* We place the page tables 16K below KERNEL_RAM_VADDR. Therefore, we must
* make sure that KERNEL_RAM_VADDR is correctly set. Currently, we expect
* the least significant 16 bits to be 0x8000, but we could probably
* relax this restriction to KERNEL_RAM_VADDR >= PAGE_OFFSET + 0x4000.
*/
#define KERNEL_RAM_VADDR (PAGE_OFFSET + TEXT_OFFSET)
#if (KERNEL_RAM_VADDR & 0xffff) != 0x8000
#error KERNEL_RAM_VADDR must start at 0xXXXX8000
#endif
#ifdef CONFIG_ARM_LPAE
/* LPAE requires an additional page for the PGD */
#define PG_DIR_SIZE 0x5000
#define PMD_ORDER 3
#else
#define PG_DIR_SIZE 0x4000
#define PMD_ORDER 2
#endif
.globl swapper_pg_dir
.equ swapper_pg_dir, KERNEL_RAM_VADDR - PG_DIR_SIZE
.macro pgtbl, rd, phys
add \rd, \phys, #TEXT_OFFSET
sub \rd, \rd, #PG_DIR_SIZE
.endm
/*
* Kernel startup entry point.
* ---------------------------
*
* This is normally called from the decompressor code. The requirements
* are: MMU = off, D-cache = off, I-cache = dont care, r0 = 0,
* r1 = machine nr, r2 = atags or dtb pointer.
*
* This code is mostly position independent, so if you link the kernel at
* 0xc0008000, you call this at __pa(0xc0008000).
*
* See linux/arch/arm/tools/mach-types for the complete list of machine
* numbers for r1.
*
* We're trying to keep crap to a minimum; DO NOT add any machine specific
* crap here - that's what the boot loader (or in extreme, well justified
* circumstances, zImage) is for.
*/
.arm
__HEAD
ENTRY(stext)
ARM_BE8(setend be ) @ ensure we are in BE8 mode
THUMB( badr r9, 1f ) @ Kernel is always entered in ARM.
THUMB( bx r9 ) @ If this is a Thumb-2 kernel,
THUMB( .thumb ) @ switch to Thumb now.
THUMB(1: )
#ifdef CONFIG_ARM_VIRT_EXT
bl __hyp_stub_install
#endif
@ ensure svc mode and all interrupts masked
safe_svcmode_maskall r9
mrc p15, 0, r9, c0, c0 @ get processor id
bl __lookup_processor_type @ r5=procinfo r9=cpuid
movs r10, r5 @ invalid processor (r5=0)?
THUMB( it eq ) @ force fixup-able long branch encoding
beq __error_p @ yes, error 'p'
#ifdef CONFIG_ARM_LPAE
mrc p15, 0, r3, c0, c1, 4 @ read ID_MMFR0
and r3, r3, #0xf @ extract VMSA support
cmp r3, #5 @ long-descriptor translation table format?
THUMB( it lo ) @ force fixup-able long branch encoding
blo __error_lpae @ only classic page table format
#endif
#ifndef CONFIG_XIP_KERNEL
adr r3, 2f
ldmia r3, {r4, r8}
sub r4, r3, r4 @ (PHYS_OFFSET - PAGE_OFFSET)
add r8, r8, r4 @ PHYS_OFFSET
#else
ldr r8, =PLAT_PHYS_OFFSET @ always constant in this case
#endif
/*
* r1 = machine no, r2 = atags or dtb,
* r8 = phys_offset, r9 = cpuid, r10 = procinfo
*/
bl __vet_atags
#ifdef CONFIG_SMP_ON_UP
bl __fixup_smp
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
#endif
#ifdef CONFIG_ARM_PATCH_PHYS_VIRT
bl __fixup_pv_table
#endif
bl __create_page_tables
/*
* The following calls CPU specific code in a position independent
* manner. See arch/arm/mm/proc-*.S for details. r10 = base of
* xxx_proc_info structure selected by __lookup_processor_type
* above.
*
* The processor init function will be called with:
* r1 - machine type
* r2 - boot data (atags/dt) pointer
* r4 - translation table base (low word)
* r5 - translation table base (high word, if LPAE)
* r8 - translation table base 1 (pfn if LPAE)
* r9 - cpuid
* r13 - virtual address for __enable_mmu -> __turn_mmu_on
*
* On return, the CPU will be ready for the MMU to be turned on,
* r0 will hold the CPU control register value, r1, r2, r4, and
* r9 will be preserved. r5 will also be preserved if LPAE.
*/
ldr r13, =__mmap_switched @ address to jump to after
@ mmu has been enabled
badr lr, 1f @ return (PIC) address
#ifdef CONFIG_ARM_LPAE
mov r5, #0 @ high TTBR0
mov r8, r4, lsr #12 @ TTBR1 is swapper_pg_dir pfn
#else
mov r8, r4 @ set TTBR1 to swapper_pg_dir
#endif
ldr r12, [r10, #PROCINFO_INITFUNC]
add r12, r12, r10
ret r12
1: b __enable_mmu
ENDPROC(stext)
.ltorg
#ifndef CONFIG_XIP_KERNEL
2: .long .
.long PAGE_OFFSET
#endif
/*
* Setup the initial page tables. We only setup the barest
* amount which are required to get the kernel running, which
* generally means mapping in the kernel code.
*
* r8 = phys_offset, r9 = cpuid, r10 = procinfo
*
* Returns:
* r0, r3, r5-r7 corrupted
* r4 = physical page table address
*/
__create_page_tables:
pgtbl r4, r8 @ page table address
/*
* Clear the swapper page table
*/
mov r0, r4
mov r3, #0
add r6, r0, #PG_DIR_SIZE
1: str r3, [r0], #4
str r3, [r0], #4
str r3, [r0], #4
str r3, [r0], #4
teq r0, r6
bne 1b
#ifdef CONFIG_ARM_LPAE
/*
* Build the PGD table (first level) to point to the PMD table. A PGD
* entry is 64-bit wide.
*/
mov r0, r4
add r3, r4, #0x1000 @ first PMD table address
orr r3, r3, #3 @ PGD block type
mov r6, #4 @ PTRS_PER_PGD
mov r7, #1 << (55 - 32) @ L_PGD_SWAPPER
1:
#ifdef CONFIG_CPU_ENDIAN_BE8
str r7, [r0], #4 @ set top PGD entry bits
str r3, [r0], #4 @ set bottom PGD entry bits
#else
str r3, [r0], #4 @ set bottom PGD entry bits
str r7, [r0], #4 @ set top PGD entry bits
#endif
add r3, r3, #0x1000 @ next PMD table
subs r6, r6, #1
bne 1b
add r4, r4, #0x1000 @ point to the PMD tables
#ifdef CONFIG_CPU_ENDIAN_BE8
add r4, r4, #4 @ we only write the bottom word
#endif
#endif
ldr r7, [r10, #PROCINFO_MM_MMUFLAGS] @ mm_mmuflags
/*
* Create identity mapping to cater for __enable_mmu.
* This identity mapping will be removed by paging_init().
*/
adr r0, __turn_mmu_on_loc
ldmia r0, {r3, r5, r6}
sub r0, r0, r3 @ virt->phys offset
add r5, r5, r0 @ phys __turn_mmu_on
add r6, r6, r0 @ phys __turn_mmu_on_end
mov r5, r5, lsr #SECTION_SHIFT
mov r6, r6, lsr #SECTION_SHIFT
1: orr r3, r7, r5, lsl #SECTION_SHIFT @ flags + kernel base
str r3, [r4, r5, lsl #PMD_ORDER] @ identity mapping
cmp r5, r6
addlo r5, r5, #1 @ next section
blo 1b
/*
* Map our RAM from the start to the end of the kernel .bss section.
*/
add r0, r4, #PAGE_OFFSET >> (SECTION_SHIFT - PMD_ORDER)
ldr r6, =(_end - 1)
orr r3, r8, r7
add r6, r4, r6, lsr #(SECTION_SHIFT - PMD_ORDER)
1: str r3, [r0], #1 << PMD_ORDER
add r3, r3, #1 << SECTION_SHIFT
cmp r0, r6
bls 1b
#ifdef CONFIG_XIP_KERNEL
/*
* Map the kernel image separately as it is not located in RAM.
*/
#define XIP_START XIP_VIRT_ADDR(CONFIG_XIP_PHYS_ADDR)
mov r3, pc
mov r3, r3, lsr #SECTION_SHIFT
orr r3, r7, r3, lsl #SECTION_SHIFT
add r0, r4, #(XIP_START & 0xff000000) >> (SECTION_SHIFT - PMD_ORDER)
str r3, [r0, #((XIP_START & 0x00f00000) >> SECTION_SHIFT) << PMD_ORDER]!
ldr r6, =(_edata_loc - 1)
add r0, r0, #1 << PMD_ORDER
add r6, r4, r6, lsr #(SECTION_SHIFT - PMD_ORDER)
1: cmp r0, r6
add r3, r3, #1 << SECTION_SHIFT
strls r3, [r0], #1 << PMD_ORDER
bls 1b
#endif
/*
* Then map boot params address in r2 if specified.
* We map 2 sections in case the ATAGs/DTB crosses a section boundary.
*/
mov r0, r2, lsr #SECTION_SHIFT
movs r0, r0, lsl #SECTION_SHIFT
subne r3, r0, r8
addne r3, r3, #PAGE_OFFSET
addne r3, r4, r3, lsr #(SECTION_SHIFT - PMD_ORDER)
orrne r6, r7, r0
strne r6, [r3], #1 << PMD_ORDER
addne r6, r6, #1 << SECTION_SHIFT
strne r6, [r3]
#if defined(CONFIG_ARM_LPAE) && defined(CONFIG_CPU_ENDIAN_BE8)
sub r4, r4, #4 @ Fixup page table pointer
@ for 64-bit descriptors
#endif
#ifdef CONFIG_DEBUG_LL
#if !defined(CONFIG_DEBUG_ICEDCC) && !defined(CONFIG_DEBUG_SEMIHOSTING)
/*
* Map in IO space for serial debugging.
* This allows debug messages to be output
* via a serial console before paging_init.
*/
addruart r7, r3, r0
mov r3, r3, lsr #SECTION_SHIFT
mov r3, r3, lsl #PMD_ORDER
add r0, r4, r3
mov r3, r7, lsr #SECTION_SHIFT
ldr r7, [r10, #PROCINFO_IO_MMUFLAGS] @ io_mmuflags
orr r3, r7, r3, lsl #SECTION_SHIFT
#ifdef CONFIG_ARM_LPAE
mov r7, #1 << (54 - 32) @ XN
#ifdef CONFIG_CPU_ENDIAN_BE8
str r7, [r0], #4
str r3, [r0], #4
#else
str r3, [r0], #4
str r7, [r0], #4
#endif
#else
orr r3, r3, #PMD_SECT_XN
str r3, [r0], #4
#endif
#else /* CONFIG_DEBUG_ICEDCC || CONFIG_DEBUG_SEMIHOSTING */
/* we don't need any serial debugging mappings */
ldr r7, [r10, #PROCINFO_IO_MMUFLAGS] @ io_mmuflags
#endif
#if defined(CONFIG_ARCH_NETWINDER) || defined(CONFIG_ARCH_CATS)
/*
* If we're using the NetWinder or CATS, we also need to map
* in the 16550-type serial port for the debug messages
*/
add r0, r4, #0xff000000 >> (SECTION_SHIFT - PMD_ORDER)
orr r3, r7, #0x7c000000
str r3, [r0]
#endif
#ifdef CONFIG_ARCH_RPC
/*
* Map in screen at 0x02000000 & SCREEN2_BASE
* Similar reasons here - for debug. This is
* only for Acorn RiscPC architectures.
*/
add r0, r4, #0x02000000 >> (SECTION_SHIFT - PMD_ORDER)
orr r3, r7, #0x02000000
str r3, [r0]
add r0, r4, #0xd8000000 >> (SECTION_SHIFT - PMD_ORDER)
str r3, [r0]
#endif
#endif
#ifdef CONFIG_ARM_LPAE
sub r4, r4, #0x1000 @ point to the PGD table
#endif
ARM: convert all "mov.* pc, reg" to "bx reg" for ARMv6+ ARMv6 and greater introduced a new instruction ("bx") which can be used to return from function calls. Recent CPUs perform better when the "bx lr" instruction is used rather than the "mov pc, lr" instruction, and this sequence is strongly recommended to be used by the ARM architecture manual (section A.4.1.1). We provide a new macro "ret" with all its variants for the condition code which will resolve to the appropriate instruction. Rather than doing this piecemeal, and miss some instances, change all the "mov pc" instances to use the new macro, with the exception of the "movs" instruction and the kprobes code. This allows us to detect the "mov pc, lr" case and fix it up - and also gives us the possibility of deploying this for other registers depending on the CPU selection. Reported-by: Will Deacon <will.deacon@arm.com> Tested-by: Stephen Warren <swarren@nvidia.com> # Tegra Jetson TK1 Tested-by: Robert Jarzmik <robert.jarzmik@free.fr> # mioa701_bootresume.S Tested-by: Andrew Lunn <andrew@lunn.ch> # Kirkwood Tested-by: Shawn Guo <shawn.guo@freescale.com> Tested-by: Tony Lindgren <tony@atomide.com> # OMAPs Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com> # Armada XP, 375, 385 Acked-by: Sekhar Nori <nsekhar@ti.com> # DaVinci Acked-by: Christoffer Dall <christoffer.dall@linaro.org> # kvm/hyp Acked-by: Haojian Zhuang <haojian.zhuang@gmail.com> # PXA3xx Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> # Xen Tested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> # ARMv7M Tested-by: Simon Horman <horms+renesas@verge.net.au> # Shmobile Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-06-30 22:29:12 +07:00
ret lr
ENDPROC(__create_page_tables)
.ltorg
.align
__turn_mmu_on_loc:
.long .
.long __turn_mmu_on
.long __turn_mmu_on_end
#if defined(CONFIG_SMP)
.text
.arm
ENTRY(secondary_startup_arm)
THUMB( badr r9, 1f ) @ Kernel is entered in ARM.
THUMB( bx r9 ) @ If this is a Thumb-2 kernel,
THUMB( .thumb ) @ switch to Thumb now.
THUMB(1: )
ENTRY(secondary_startup)
/*
* Common entry point for secondary CPUs.
*
* Ensure that we're in SVC mode, and IRQs are disabled. Lookup
* the processor type - there is no need to check the machine type
* as it has already been validated by the primary processor.
*/
ARM_BE8(setend be) @ ensure we are in BE8 mode
#ifdef CONFIG_ARM_VIRT_EXT
bl __hyp_stub_install_secondary
#endif
safe_svcmode_maskall r9
mrc p15, 0, r9, c0, c0 @ get processor id
bl __lookup_processor_type
movs r10, r5 @ invalid processor?
moveq r0, #'p' @ yes, error 'p'
THUMB( it eq ) @ force fixup-able long branch encoding
beq __error_p
/*
* Use the page tables supplied from __cpu_up.
*/
adr r4, __secondary_data
ldmia r4, {r5, r7, r12} @ address to jump to after
sub lr, r4, r5 @ mmu has been enabled
add r3, r7, lr
ldrd r4, [r3, #0] @ get secondary_data.pgdir
ARM_BE8(eor r4, r4, r5) @ Swap r5 and r4 in BE:
ARM_BE8(eor r5, r4, r5) @ it can be done in 3 steps
ARM_BE8(eor r4, r4, r5) @ without using a temp reg.
ldr r8, [r3, #8] @ get secondary_data.swapper_pg_dir
badr lr, __enable_mmu @ return address
mov r13, r12 @ __secondary_switched address
ldr r12, [r10, #PROCINFO_INITFUNC]
add r12, r12, r10 @ initialise processor
@ (return control reg)
ret r12
ENDPROC(secondary_startup)
ENDPROC(secondary_startup_arm)
/*
* r6 = &secondary_data
*/
ENTRY(__secondary_switched)
ldr sp, [r7, #12] @ get secondary_data.stack
mov fp, #0
b secondary_start_kernel
ENDPROC(__secondary_switched)
.align
.type __secondary_data, %object
__secondary_data:
.long .
.long secondary_data
.long __secondary_switched
#endif /* defined(CONFIG_SMP) */
/*
* Setup common bits before finally enabling the MMU. Essentially
* this is just loading the page table pointer and domain access
* registers. All these registers need to be preserved by the
* processor setup function (or set in the case of r0)
*
* r0 = cp#15 control register
* r1 = machine ID
* r2 = atags or dtb pointer
* r4 = TTBR pointer (low word)
* r5 = TTBR pointer (high word if LPAE)
* r9 = processor ID
* r13 = *virtual* address to jump to upon completion
*/
__enable_mmu:
#if defined(CONFIG_ALIGNMENT_TRAP) && __LINUX_ARM_ARCH__ < 6
orr r0, r0, #CR_A
#else
bic r0, r0, #CR_A
#endif
#ifdef CONFIG_CPU_DCACHE_DISABLE
bic r0, r0, #CR_C
#endif
#ifdef CONFIG_CPU_BPREDICT_DISABLE
bic r0, r0, #CR_Z
#endif
#ifdef CONFIG_CPU_ICACHE_DISABLE
bic r0, r0, #CR_I
#endif
#ifdef CONFIG_ARM_LPAE
mcrr p15, 0, r4, r5, c2 @ load TTBR0
#else
mov r5, #DACR_INIT
mcr p15, 0, r5, c3, c0, 0 @ load domain access register
mcr p15, 0, r4, c2, c0, 0 @ load page table pointer
#endif
b __turn_mmu_on
ENDPROC(__enable_mmu)
/*
* Enable the MMU. This completely changes the structure of the visible
* memory space. You will not be able to trace execution through this.
* If you have an enquiry about this, *please* check the linux-arm-kernel
* mailing list archives BEFORE sending another post to the list.
*
* r0 = cp#15 control register
* r1 = machine ID
* r2 = atags or dtb pointer
* r9 = processor ID
* r13 = *virtual* address to jump to upon completion
*
* other registers depend on the function called upon completion
*/
.align 5
.pushsection .idmap.text, "ax"
ENTRY(__turn_mmu_on)
mov r0, r0
instr_sync
mcr p15, 0, r0, c1, c0, 0 @ write control reg
mrc p15, 0, r3, c0, c0, 0 @ read id reg
instr_sync
mov r3, r3
mov r3, r13
ARM: convert all "mov.* pc, reg" to "bx reg" for ARMv6+ ARMv6 and greater introduced a new instruction ("bx") which can be used to return from function calls. Recent CPUs perform better when the "bx lr" instruction is used rather than the "mov pc, lr" instruction, and this sequence is strongly recommended to be used by the ARM architecture manual (section A.4.1.1). We provide a new macro "ret" with all its variants for the condition code which will resolve to the appropriate instruction. Rather than doing this piecemeal, and miss some instances, change all the "mov pc" instances to use the new macro, with the exception of the "movs" instruction and the kprobes code. This allows us to detect the "mov pc, lr" case and fix it up - and also gives us the possibility of deploying this for other registers depending on the CPU selection. Reported-by: Will Deacon <will.deacon@arm.com> Tested-by: Stephen Warren <swarren@nvidia.com> # Tegra Jetson TK1 Tested-by: Robert Jarzmik <robert.jarzmik@free.fr> # mioa701_bootresume.S Tested-by: Andrew Lunn <andrew@lunn.ch> # Kirkwood Tested-by: Shawn Guo <shawn.guo@freescale.com> Tested-by: Tony Lindgren <tony@atomide.com> # OMAPs Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com> # Armada XP, 375, 385 Acked-by: Sekhar Nori <nsekhar@ti.com> # DaVinci Acked-by: Christoffer Dall <christoffer.dall@linaro.org> # kvm/hyp Acked-by: Haojian Zhuang <haojian.zhuang@gmail.com> # PXA3xx Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> # Xen Tested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> # ARMv7M Tested-by: Simon Horman <horms+renesas@verge.net.au> # Shmobile Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-06-30 22:29:12 +07:00
ret r3
__turn_mmu_on_end:
ENDPROC(__turn_mmu_on)
.popsection
#ifdef CONFIG_SMP_ON_UP
__HEAD
__fixup_smp:
and r3, r9, #0x000f0000 @ architecture version
teq r3, #0x000f0000 @ CPU ID supported?
bne __fixup_smp_on_up @ no, assume UP
bic r3, r9, #0x00ff0000
bic r3, r3, #0x0000000f @ mask 0xff00fff0
mov r4, #0x41000000
orr r4, r4, #0x0000b000
orr r4, r4, #0x00000020 @ val 0x4100b020
teq r3, r4 @ ARM 11MPCore?
ARM: convert all "mov.* pc, reg" to "bx reg" for ARMv6+ ARMv6 and greater introduced a new instruction ("bx") which can be used to return from function calls. Recent CPUs perform better when the "bx lr" instruction is used rather than the "mov pc, lr" instruction, and this sequence is strongly recommended to be used by the ARM architecture manual (section A.4.1.1). We provide a new macro "ret" with all its variants for the condition code which will resolve to the appropriate instruction. Rather than doing this piecemeal, and miss some instances, change all the "mov pc" instances to use the new macro, with the exception of the "movs" instruction and the kprobes code. This allows us to detect the "mov pc, lr" case and fix it up - and also gives us the possibility of deploying this for other registers depending on the CPU selection. Reported-by: Will Deacon <will.deacon@arm.com> Tested-by: Stephen Warren <swarren@nvidia.com> # Tegra Jetson TK1 Tested-by: Robert Jarzmik <robert.jarzmik@free.fr> # mioa701_bootresume.S Tested-by: Andrew Lunn <andrew@lunn.ch> # Kirkwood Tested-by: Shawn Guo <shawn.guo@freescale.com> Tested-by: Tony Lindgren <tony@atomide.com> # OMAPs Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com> # Armada XP, 375, 385 Acked-by: Sekhar Nori <nsekhar@ti.com> # DaVinci Acked-by: Christoffer Dall <christoffer.dall@linaro.org> # kvm/hyp Acked-by: Haojian Zhuang <haojian.zhuang@gmail.com> # PXA3xx Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> # Xen Tested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> # ARMv7M Tested-by: Simon Horman <horms+renesas@verge.net.au> # Shmobile Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-06-30 22:29:12 +07:00
reteq lr @ yes, assume SMP
mrc p15, 0, r0, c0, c0, 5 @ read MPIDR
and r0, r0, #0xc0000000 @ multiprocessing extensions and
teq r0, #0x80000000 @ not part of a uniprocessor system?
bne __fixup_smp_on_up @ no, assume UP
@ Core indicates it is SMP. Check for Aegis SOC where a single
@ Cortex-A9 CPU is present but SMP operations fault.
mov r4, #0x41000000
orr r4, r4, #0x0000c000
orr r4, r4, #0x00000090
teq r3, r4 @ Check for ARM Cortex-A9
ARM: convert all "mov.* pc, reg" to "bx reg" for ARMv6+ ARMv6 and greater introduced a new instruction ("bx") which can be used to return from function calls. Recent CPUs perform better when the "bx lr" instruction is used rather than the "mov pc, lr" instruction, and this sequence is strongly recommended to be used by the ARM architecture manual (section A.4.1.1). We provide a new macro "ret" with all its variants for the condition code which will resolve to the appropriate instruction. Rather than doing this piecemeal, and miss some instances, change all the "mov pc" instances to use the new macro, with the exception of the "movs" instruction and the kprobes code. This allows us to detect the "mov pc, lr" case and fix it up - and also gives us the possibility of deploying this for other registers depending on the CPU selection. Reported-by: Will Deacon <will.deacon@arm.com> Tested-by: Stephen Warren <swarren@nvidia.com> # Tegra Jetson TK1 Tested-by: Robert Jarzmik <robert.jarzmik@free.fr> # mioa701_bootresume.S Tested-by: Andrew Lunn <andrew@lunn.ch> # Kirkwood Tested-by: Shawn Guo <shawn.guo@freescale.com> Tested-by: Tony Lindgren <tony@atomide.com> # OMAPs Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com> # Armada XP, 375, 385 Acked-by: Sekhar Nori <nsekhar@ti.com> # DaVinci Acked-by: Christoffer Dall <christoffer.dall@linaro.org> # kvm/hyp Acked-by: Haojian Zhuang <haojian.zhuang@gmail.com> # PXA3xx Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> # Xen Tested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> # ARMv7M Tested-by: Simon Horman <horms+renesas@verge.net.au> # Shmobile Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-06-30 22:29:12 +07:00
retne lr @ Not ARM Cortex-A9,
@ If a future SoC *does* use 0x0 as the PERIPH_BASE, then the
@ below address check will need to be #ifdef'd or equivalent
@ for the Aegis platform.
mrc p15, 4, r0, c15, c0 @ get SCU base address
teq r0, #0x0 @ '0' on actual UP A9 hardware
beq __fixup_smp_on_up @ So its an A9 UP
ldr r0, [r0, #4] @ read SCU Config
ARM_BE8(rev r0, r0) @ byteswap if big endian
and r0, r0, #0x3 @ number of CPUs
teq r0, #0x0 @ is 1?
ARM: convert all "mov.* pc, reg" to "bx reg" for ARMv6+ ARMv6 and greater introduced a new instruction ("bx") which can be used to return from function calls. Recent CPUs perform better when the "bx lr" instruction is used rather than the "mov pc, lr" instruction, and this sequence is strongly recommended to be used by the ARM architecture manual (section A.4.1.1). We provide a new macro "ret" with all its variants for the condition code which will resolve to the appropriate instruction. Rather than doing this piecemeal, and miss some instances, change all the "mov pc" instances to use the new macro, with the exception of the "movs" instruction and the kprobes code. This allows us to detect the "mov pc, lr" case and fix it up - and also gives us the possibility of deploying this for other registers depending on the CPU selection. Reported-by: Will Deacon <will.deacon@arm.com> Tested-by: Stephen Warren <swarren@nvidia.com> # Tegra Jetson TK1 Tested-by: Robert Jarzmik <robert.jarzmik@free.fr> # mioa701_bootresume.S Tested-by: Andrew Lunn <andrew@lunn.ch> # Kirkwood Tested-by: Shawn Guo <shawn.guo@freescale.com> Tested-by: Tony Lindgren <tony@atomide.com> # OMAPs Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com> # Armada XP, 375, 385 Acked-by: Sekhar Nori <nsekhar@ti.com> # DaVinci Acked-by: Christoffer Dall <christoffer.dall@linaro.org> # kvm/hyp Acked-by: Haojian Zhuang <haojian.zhuang@gmail.com> # PXA3xx Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> # Xen Tested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> # ARMv7M Tested-by: Simon Horman <horms+renesas@verge.net.au> # Shmobile Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-06-30 22:29:12 +07:00
retne lr
__fixup_smp_on_up:
adr r0, 1f
ldmia r0, {r3 - r5}
sub r3, r0, r3
add r4, r4, r3
add r5, r5, r3
b __do_fixup_smp_on_up
ENDPROC(__fixup_smp)
.align
1: .word .
.word __smpalt_begin
.word __smpalt_end
.pushsection .data
.align 2
.globl smp_on_up
smp_on_up:
ALT_SMP(.long 1)
ALT_UP(.long 0)
.popsection
#endif
.text
__do_fixup_smp_on_up:
cmp r4, r5
ARM: convert all "mov.* pc, reg" to "bx reg" for ARMv6+ ARMv6 and greater introduced a new instruction ("bx") which can be used to return from function calls. Recent CPUs perform better when the "bx lr" instruction is used rather than the "mov pc, lr" instruction, and this sequence is strongly recommended to be used by the ARM architecture manual (section A.4.1.1). We provide a new macro "ret" with all its variants for the condition code which will resolve to the appropriate instruction. Rather than doing this piecemeal, and miss some instances, change all the "mov pc" instances to use the new macro, with the exception of the "movs" instruction and the kprobes code. This allows us to detect the "mov pc, lr" case and fix it up - and also gives us the possibility of deploying this for other registers depending on the CPU selection. Reported-by: Will Deacon <will.deacon@arm.com> Tested-by: Stephen Warren <swarren@nvidia.com> # Tegra Jetson TK1 Tested-by: Robert Jarzmik <robert.jarzmik@free.fr> # mioa701_bootresume.S Tested-by: Andrew Lunn <andrew@lunn.ch> # Kirkwood Tested-by: Shawn Guo <shawn.guo@freescale.com> Tested-by: Tony Lindgren <tony@atomide.com> # OMAPs Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com> # Armada XP, 375, 385 Acked-by: Sekhar Nori <nsekhar@ti.com> # DaVinci Acked-by: Christoffer Dall <christoffer.dall@linaro.org> # kvm/hyp Acked-by: Haojian Zhuang <haojian.zhuang@gmail.com> # PXA3xx Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> # Xen Tested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> # ARMv7M Tested-by: Simon Horman <horms+renesas@verge.net.au> # Shmobile Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-06-30 22:29:12 +07:00
reths lr
ldmia r4!, {r0, r6}
ARM( str r6, [r0, r3] )
THUMB( add r0, r0, r3 )
#ifdef __ARMEB__
THUMB( mov r6, r6, ror #16 ) @ Convert word order for big-endian.
#endif
THUMB( strh r6, [r0], #2 ) @ For Thumb-2, store as two halfwords
THUMB( mov r6, r6, lsr #16 ) @ to be robust against misaligned r3.
THUMB( strh r6, [r0] )
b __do_fixup_smp_on_up
ENDPROC(__do_fixup_smp_on_up)
ENTRY(fixup_smp)
stmfd sp!, {r4 - r6, lr}
mov r4, r0
add r5, r0, r1
mov r3, #0
bl __do_fixup_smp_on_up
ldmfd sp!, {r4 - r6, pc}
ENDPROC(fixup_smp)
#ifdef __ARMEB__
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
#define LOW_OFFSET 0x4
#define HIGH_OFFSET 0x0
#else
#define LOW_OFFSET 0x0
#define HIGH_OFFSET 0x4
#endif
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
#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'.
*/
__HEAD
__fixup_pv_table:
adr r0, 1f
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
ldmia r0, {r3-r7}
mvn ip, #0
subs r3, r0, r3 @ PHYS_OFFSET - PAGE_OFFSET
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
add r4, r4, r3 @ adjust table start address
add r5, r5, r3 @ adjust table end address
ARM: Better virt_to_page() handling virt_to_page() is incredibly inefficient when virt-to-phys patching is enabled. This is because we end up with this calculation: page = &mem_map[asm virt_to_phys(addr) >> 12 - __pv_phys_offset >> 12] in assembly. The asm virt_to_phys() is equivalent this this operation: addr - PAGE_OFFSET + __pv_phys_offset and we can see that because this is assembly, the compiler has no chance to optimise some of that away. This should reduce down to: page = &mem_map[(addr - PAGE_OFFSET) >> 12] for the common cases. Permit the compiler to make this optimisation by giving it more of the information it needs - do this by providing a virt_to_pfn() macro. Another issue which makes this more complex is that __pv_phys_offset is a 64-bit type on all platforms. This is needlessly wasteful - if we store the physical offset as a PFN, we can save a lot of work having to deal with 64-bit values, which sometimes ends up producing incredibly horrid code: a4c: e3009000 movw r9, #0 a4c: R_ARM_MOVW_ABS_NC __pv_phys_offset a50: e3409000 movt r9, #0 ; r9 = &__pv_phys_offset a50: R_ARM_MOVT_ABS __pv_phys_offset a54: e3002000 movw r2, #0 a54: R_ARM_MOVW_ABS_NC __pv_phys_offset a58: e3402000 movt r2, #0 ; r2 = &__pv_phys_offset a58: R_ARM_MOVT_ABS __pv_phys_offset a5c: e5999004 ldr r9, [r9, #4] ; r9 = high word of __pv_phys_offset a60: e3001000 movw r1, #0 a60: R_ARM_MOVW_ABS_NC mem_map a64: e592c000 ldr ip, [r2] ; ip = low word of __pv_phys_offset Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-03-26 02:45:31 +07:00
add r6, r6, r3 @ adjust __pv_phys_pfn_offset address
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
add r7, r7, r3 @ adjust __pv_offset address
mov r0, r8, lsr #PAGE_SHIFT @ convert to PFN
str r0, [r6] @ save computed PHYS_OFFSET to __pv_phys_pfn_offset
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
strcc ip, [r7, #HIGH_OFFSET] @ save to __pv_offset high bits
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
mov r6, r3, lsr #24 @ constant for add/sub instructions
teq r3, r6, lsl #24 @ must be 16MiB aligned
THUMB( it ne @ cross section branch )
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
bne __error
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
str r3, [r7, #LOW_OFFSET] @ save to __pv_offset low bits
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
b __fixup_a_pv_table
ENDPROC(__fixup_pv_table)
.align
1: .long .
.long __pv_table_begin
.long __pv_table_end
ARM: Better virt_to_page() handling virt_to_page() is incredibly inefficient when virt-to-phys patching is enabled. This is because we end up with this calculation: page = &mem_map[asm virt_to_phys(addr) >> 12 - __pv_phys_offset >> 12] in assembly. The asm virt_to_phys() is equivalent this this operation: addr - PAGE_OFFSET + __pv_phys_offset and we can see that because this is assembly, the compiler has no chance to optimise some of that away. This should reduce down to: page = &mem_map[(addr - PAGE_OFFSET) >> 12] for the common cases. Permit the compiler to make this optimisation by giving it more of the information it needs - do this by providing a virt_to_pfn() macro. Another issue which makes this more complex is that __pv_phys_offset is a 64-bit type on all platforms. This is needlessly wasteful - if we store the physical offset as a PFN, we can save a lot of work having to deal with 64-bit values, which sometimes ends up producing incredibly horrid code: a4c: e3009000 movw r9, #0 a4c: R_ARM_MOVW_ABS_NC __pv_phys_offset a50: e3409000 movt r9, #0 ; r9 = &__pv_phys_offset a50: R_ARM_MOVT_ABS __pv_phys_offset a54: e3002000 movw r2, #0 a54: R_ARM_MOVW_ABS_NC __pv_phys_offset a58: e3402000 movt r2, #0 ; r2 = &__pv_phys_offset a58: R_ARM_MOVT_ABS __pv_phys_offset a5c: e5999004 ldr r9, [r9, #4] ; r9 = high word of __pv_phys_offset a60: e3001000 movw r1, #0 a60: R_ARM_MOVW_ABS_NC mem_map a64: e592c000 ldr ip, [r2] ; ip = low word of __pv_phys_offset Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-03-26 02:45:31 +07:00
2: .long __pv_phys_pfn_offset
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
.long __pv_offset
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
.text
__fixup_a_pv_table:
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
adr r0, 3f
ldr r6, [r0]
add r6, r6, r3
ldr r0, [r6, #HIGH_OFFSET] @ pv_offset high word
ldr r6, [r6, #LOW_OFFSET] @ pv_offset low word
mov r6, r6, lsr #24
cmn r0, #1
#ifdef CONFIG_THUMB2_KERNEL
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
moveq r0, #0x200000 @ set bit 21, mov to mvn instruction
lsls r6, #24
beq 2f
clz r7, r6
lsr r6, #24
lsl r6, r7
bic r6, #0x0080
lsrs r7, #1
orrcs r6, #0x0080
orr r6, r6, r7, lsl #12
orr r6, #0x4000
b 2f
1: add r7, r3
ldrh ip, [r7, #2]
ARM_BE8(rev16 ip, ip)
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
tst ip, #0x4000
and ip, #0x8f00
orrne ip, r6 @ mask in offset bits 31-24
orreq ip, r0 @ mask in offset bits 7-0
ARM_BE8(rev16 ip, ip)
strh ip, [r7, #2]
bne 2f
ldrh ip, [r7]
ARM_BE8(rev16 ip, ip)
bic ip, #0x20
orr ip, ip, r0, lsr #16
ARM_BE8(rev16 ip, ip)
strh ip, [r7]
2: cmp r4, r5
ldrcc r7, [r4], #4 @ use branch for delay slot
bcc 1b
bx lr
#else
#ifdef CONFIG_CPU_ENDIAN_BE8
moveq r0, #0x00004000 @ set bit 22, mov to mvn instruction
#else
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
moveq r0, #0x400000 @ set bit 22, mov to mvn instruction
#endif
b 2f
1: ldr ip, [r7, r3]
#ifdef CONFIG_CPU_ENDIAN_BE8
@ in BE8, we load data in BE, but instructions still in LE
bic ip, ip, #0xff000000
tst ip, #0x000f0000 @ check the rotation field
orrne ip, ip, r6, lsl #24 @ mask in offset bits 31-24
biceq ip, ip, #0x00004000 @ clear bit 22
orreq ip, ip, r0 @ mask in offset bits 7-0
#else
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
bic ip, ip, #0x000000ff
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
tst ip, #0xf00 @ check the rotation field
orrne ip, ip, r6 @ mask in offset bits 31-24
biceq ip, ip, #0x400000 @ clear bit 22
orreq ip, ip, r0 @ mask in offset bits 7-0
#endif
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
str ip, [r7, r3]
2: cmp r4, r5
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
ldrcc r7, [r4], #4 @ use branch for delay slot
bcc 1b
ARM: convert all "mov.* pc, reg" to "bx reg" for ARMv6+ ARMv6 and greater introduced a new instruction ("bx") which can be used to return from function calls. Recent CPUs perform better when the "bx lr" instruction is used rather than the "mov pc, lr" instruction, and this sequence is strongly recommended to be used by the ARM architecture manual (section A.4.1.1). We provide a new macro "ret" with all its variants for the condition code which will resolve to the appropriate instruction. Rather than doing this piecemeal, and miss some instances, change all the "mov pc" instances to use the new macro, with the exception of the "movs" instruction and the kprobes code. This allows us to detect the "mov pc, lr" case and fix it up - and also gives us the possibility of deploying this for other registers depending on the CPU selection. Reported-by: Will Deacon <will.deacon@arm.com> Tested-by: Stephen Warren <swarren@nvidia.com> # Tegra Jetson TK1 Tested-by: Robert Jarzmik <robert.jarzmik@free.fr> # mioa701_bootresume.S Tested-by: Andrew Lunn <andrew@lunn.ch> # Kirkwood Tested-by: Shawn Guo <shawn.guo@freescale.com> Tested-by: Tony Lindgren <tony@atomide.com> # OMAPs Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com> # Armada XP, 375, 385 Acked-by: Sekhar Nori <nsekhar@ti.com> # DaVinci Acked-by: Christoffer Dall <christoffer.dall@linaro.org> # kvm/hyp Acked-by: Haojian Zhuang <haojian.zhuang@gmail.com> # PXA3xx Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com> # Xen Tested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> # ARMv7M Tested-by: Simon Horman <horms+renesas@verge.net.au> # Shmobile Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-06-30 22:29:12 +07:00
ret lr
#endif
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
ENDPROC(__fixup_a_pv_table)
.align
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
3: .long __pv_offset
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
ENTRY(fixup_pv_table)
stmfd sp!, {r4 - r7, lr}
mov r3, #0 @ no offset
mov r4, r0 @ r0 = table start
add r5, r0, r1 @ r1 = table size
bl __fixup_a_pv_table
ldmfd sp!, {r4 - r7, pc}
ENDPROC(fixup_pv_table)
.data
.align 2
ARM: Better virt_to_page() handling virt_to_page() is incredibly inefficient when virt-to-phys patching is enabled. This is because we end up with this calculation: page = &mem_map[asm virt_to_phys(addr) >> 12 - __pv_phys_offset >> 12] in assembly. The asm virt_to_phys() is equivalent this this operation: addr - PAGE_OFFSET + __pv_phys_offset and we can see that because this is assembly, the compiler has no chance to optimise some of that away. This should reduce down to: page = &mem_map[(addr - PAGE_OFFSET) >> 12] for the common cases. Permit the compiler to make this optimisation by giving it more of the information it needs - do this by providing a virt_to_pfn() macro. Another issue which makes this more complex is that __pv_phys_offset is a 64-bit type on all platforms. This is needlessly wasteful - if we store the physical offset as a PFN, we can save a lot of work having to deal with 64-bit values, which sometimes ends up producing incredibly horrid code: a4c: e3009000 movw r9, #0 a4c: R_ARM_MOVW_ABS_NC __pv_phys_offset a50: e3409000 movt r9, #0 ; r9 = &__pv_phys_offset a50: R_ARM_MOVT_ABS __pv_phys_offset a54: e3002000 movw r2, #0 a54: R_ARM_MOVW_ABS_NC __pv_phys_offset a58: e3402000 movt r2, #0 ; r2 = &__pv_phys_offset a58: R_ARM_MOVT_ABS __pv_phys_offset a5c: e5999004 ldr r9, [r9, #4] ; r9 = high word of __pv_phys_offset a60: e3001000 movw r1, #0 a60: R_ARM_MOVW_ABS_NC mem_map a64: e592c000 ldr ip, [r2] ; ip = low word of __pv_phys_offset Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-03-26 02:45:31 +07:00
.globl __pv_phys_pfn_offset
.type __pv_phys_pfn_offset, %object
__pv_phys_pfn_offset:
.word 0
.size __pv_phys_pfn_offset, . -__pv_phys_pfn_offset
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
.globl __pv_offset
.type __pv_offset, %object
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
__pv_offset:
ARM: mm: Correct virt_to_phys patching for 64 bit physical addresses The current phys_to_virt patching mechanism works only for 32 bit physical addresses and this patch extends the idea for 64bit physical addresses. The 64bit v2p patching mechanism patches the higher 8 bits of physical address with a constant using 'mov' instruction and lower 32bits are patched using 'add'. While this is correct, in those platforms where the lowmem addressable physical memory spawns across 4GB boundary, a carry bit can be produced as a result of addition of lower 32bits. This has to be taken in to account and added in to the upper. The patched __pv_offset and va are added in lower 32bits, where __pv_offset can be in two's complement form when PA_START < VA_START and that can result in a false carry bit. e.g 1) PA = 0x80000000; VA = 0xC0000000 __pv_offset = PA - VA = 0xC0000000 (2's complement) 2) PA = 0x2 80000000; VA = 0xC000000 __pv_offset = PA - VA = 0x1 C0000000 So adding __pv_offset + VA should never result in a true overflow for (1). So in order to differentiate between a true carry, a __pv_offset is extended to 64bit and the upper 32bits will have 0xffffffff if __pv_offset is 2's complement. So 'mvn #0' is inserted instead of 'mov' while patching for the same reason. Since mov, add, sub instruction are to patched with different constants inside the same stub, the rotation field of the opcode is using to differentiate between them. So the above examples for v2p translation becomes for VA=0xC0000000, 1) PA[63:32] = 0xffffffff PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x0 80000000 2) PA[63:32] = 0x1 PA[31:0] = VA + 0xC0000000 --> results in a carry PA[63:32] = PA[63:32] + carry PA[63:0] = 0x2 80000000 The above ideas were suggested by Nicolas Pitre <nico@linaro.org> as part of the review of first and second versions of the subject patch. There is no corresponding change on the phys_to_virt() side, because computations on the upper 32-bits would be discarded anyway. Cc: Russell King <linux@arm.linux.org.uk> Reviewed-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Sricharan R <r.sricharan@ti.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
2013-07-29 21:56:22 +07:00
.quad 0
.size __pv_offset, . -__pv_offset
ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching 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 allows us to achieve this with minimal loss of performance. 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 calculate the real 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. Add a module version magic string for this feature to prevent incompatible modules being loaded. Tested-by: Tony Lindgren <tony@atomide.com> Reviewed-by: Nicolas Pitre <nicolas.pitre@linaro.org> Tested-by: Nicolas Pitre <nicolas.pitre@linaro.org> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-01-05 02:09:43 +07:00
#endif
#include "head-common.S"