linux_dsm_epyc7002/arch/arm/mach-imx/suspend-imx6.S

350 lines
8.3 KiB
ArmAsm
Raw Normal View History

/*
* Copyright 2014 Freescale Semiconductor, Inc.
*
* The code contained herein is licensed under the GNU General Public
* License. You may obtain a copy of the GNU General Public License
* Version 2 or later at the following locations:
*
* http://www.opensource.org/licenses/gpl-license.html
* http://www.gnu.org/copyleft/gpl.html
*/
#include <linux/linkage.h>
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
#include <asm/assembler.h>
#include <asm/asm-offsets.h>
#include <asm/hardware/cache-l2x0.h>
#include "hardware.h"
/*
* ==================== low level suspend ====================
*
* Better to follow below rules to use ARM registers:
* r0: pm_info structure address;
* r1 ~ r4: for saving pm_info members;
* r5 ~ r10: free registers;
* r11: io base address.
*
* suspend ocram space layout:
* ======================== high address ======================
* .
* .
* .
* ^
* ^
* ^
* imx6_suspend code
* PM_INFO structure(imx6_cpu_pm_info)
* ======================== low address =======================
*/
/*
* Below offsets are based on struct imx6_cpu_pm_info
* which defined in arch/arm/mach-imx/pm-imx6q.c, this
* structure contains necessary pm info for low level
* suspend related code.
*/
#define PM_INFO_PBASE_OFFSET 0x0
#define PM_INFO_RESUME_ADDR_OFFSET 0x4
#define PM_INFO_DDR_TYPE_OFFSET 0x8
#define PM_INFO_PM_INFO_SIZE_OFFSET 0xC
#define PM_INFO_MX6Q_MMDC_P_OFFSET 0x10
#define PM_INFO_MX6Q_MMDC_V_OFFSET 0x14
#define PM_INFO_MX6Q_SRC_P_OFFSET 0x18
#define PM_INFO_MX6Q_SRC_V_OFFSET 0x1C
#define PM_INFO_MX6Q_IOMUXC_P_OFFSET 0x20
#define PM_INFO_MX6Q_IOMUXC_V_OFFSET 0x24
#define PM_INFO_MX6Q_CCM_P_OFFSET 0x28
#define PM_INFO_MX6Q_CCM_V_OFFSET 0x2C
#define PM_INFO_MX6Q_GPC_P_OFFSET 0x30
#define PM_INFO_MX6Q_GPC_V_OFFSET 0x34
#define PM_INFO_MX6Q_L2_P_OFFSET 0x38
#define PM_INFO_MX6Q_L2_V_OFFSET 0x3C
#define PM_INFO_MMDC_IO_NUM_OFFSET 0x40
#define PM_INFO_MMDC_IO_VAL_OFFSET 0x44
#define MX6Q_SRC_GPR1 0x20
#define MX6Q_SRC_GPR2 0x24
#define MX6Q_MMDC_MAPSR 0x404
#define MX6Q_MMDC_MPDGCTRL0 0x83c
#define MX6Q_GPC_IMR1 0x08
#define MX6Q_GPC_IMR2 0x0c
#define MX6Q_GPC_IMR3 0x10
#define MX6Q_GPC_IMR4 0x14
#define MX6Q_CCM_CCR 0x0
.align 3
.macro sync_l2_cache
/* sync L2 cache to drain L2's buffers to DRAM. */
#ifdef CONFIG_CACHE_L2X0
ldr r11, [r0, #PM_INFO_MX6Q_L2_V_OFFSET]
teq r11, #0
beq 6f
mov r6, #0x0
str r6, [r11, #L2X0_CACHE_SYNC]
1:
ldr r6, [r11, #L2X0_CACHE_SYNC]
ands r6, r6, #0x1
bne 1b
6:
#endif
.endm
.macro resume_mmdc
/* restore MMDC IO */
cmp r5, #0x0
ldreq r11, [r0, #PM_INFO_MX6Q_IOMUXC_V_OFFSET]
ldrne r11, [r0, #PM_INFO_MX6Q_IOMUXC_P_OFFSET]
ldr r6, [r0, #PM_INFO_MMDC_IO_NUM_OFFSET]
ldr r7, =PM_INFO_MMDC_IO_VAL_OFFSET
add r7, r7, r0
1:
ldr r8, [r7], #0x4
ldr r9, [r7], #0x4
str r9, [r11, r8]
subs r6, r6, #0x1
bne 1b
cmp r5, #0x0
ldreq r11, [r0, #PM_INFO_MX6Q_MMDC_V_OFFSET]
ldrne r11, [r0, #PM_INFO_MX6Q_MMDC_P_OFFSET]
cmp r3, #IMX_DDR_TYPE_LPDDR2
bne 4f
/* reset read FIFO, RST_RD_FIFO */
ldr r7, =MX6Q_MMDC_MPDGCTRL0
ldr r6, [r11, r7]
orr r6, r6, #(1 << 31)
str r6, [r11, r7]
2:
ldr r6, [r11, r7]
ands r6, r6, #(1 << 31)
bne 2b
/* reset FIFO a second time */
ldr r6, [r11, r7]
orr r6, r6, #(1 << 31)
str r6, [r11, r7]
3:
ldr r6, [r11, r7]
ands r6, r6, #(1 << 31)
bne 3b
4:
/* let DDR out of self-refresh */
ldr r7, [r11, #MX6Q_MMDC_MAPSR]
bic r7, r7, #(1 << 21)
str r7, [r11, #MX6Q_MMDC_MAPSR]
5:
ldr r7, [r11, #MX6Q_MMDC_MAPSR]
ands r7, r7, #(1 << 25)
bne 5b
/* enable DDR auto power saving */
ldr r7, [r11, #MX6Q_MMDC_MAPSR]
bic r7, r7, #0x1
str r7, [r11, #MX6Q_MMDC_MAPSR]
.endm
ENTRY(imx6_suspend)
ldr r1, [r0, #PM_INFO_PBASE_OFFSET]
ldr r2, [r0, #PM_INFO_RESUME_ADDR_OFFSET]
ldr r3, [r0, #PM_INFO_DDR_TYPE_OFFSET]
ldr r4, [r0, #PM_INFO_PM_INFO_SIZE_OFFSET]
/*
* counting the resume address in iram
* to set it in SRC register.
*/
ldr r6, =imx6_suspend
ldr r7, =resume
sub r7, r7, r6
add r8, r1, r4
add r9, r8, r7
/*
* make sure TLB contain the addr we want,
* as we will access them after MMDC IO floated.
*/
ldr r11, [r0, #PM_INFO_MX6Q_CCM_V_OFFSET]
ldr r6, [r11, #0x0]
ldr r11, [r0, #PM_INFO_MX6Q_GPC_V_OFFSET]
ldr r6, [r11, #0x0]
ldr r11, [r0, #PM_INFO_MX6Q_IOMUXC_V_OFFSET]
ldr r6, [r11, #0x0]
/* use r11 to store the IO address */
ldr r11, [r0, #PM_INFO_MX6Q_SRC_V_OFFSET]
/* store physical resume addr and pm_info address. */
str r9, [r11, #MX6Q_SRC_GPR1]
str r1, [r11, #MX6Q_SRC_GPR2]
/* need to sync L2 cache before DSM. */
sync_l2_cache
ldr r11, [r0, #PM_INFO_MX6Q_MMDC_V_OFFSET]
/*
* put DDR explicitly into self-refresh and
* disable automatic power savings.
*/
ldr r7, [r11, #MX6Q_MMDC_MAPSR]
orr r7, r7, #0x1
str r7, [r11, #MX6Q_MMDC_MAPSR]
/* make the DDR explicitly enter self-refresh. */
ldr r7, [r11, #MX6Q_MMDC_MAPSR]
orr r7, r7, #(1 << 21)
str r7, [r11, #MX6Q_MMDC_MAPSR]
poll_dvfs_set:
ldr r7, [r11, #MX6Q_MMDC_MAPSR]
ands r7, r7, #(1 << 25)
beq poll_dvfs_set
ldr r11, [r0, #PM_INFO_MX6Q_IOMUXC_V_OFFSET]
ldr r6, =0x0
ldr r7, [r0, #PM_INFO_MMDC_IO_NUM_OFFSET]
ldr r8, =PM_INFO_MMDC_IO_VAL_OFFSET
add r8, r8, r0
/* LPDDR2's last 3 IOs need special setting */
cmp r3, #IMX_DDR_TYPE_LPDDR2
subeq r7, r7, #0x3
set_mmdc_io_lpm:
ldr r9, [r8], #0x8
str r6, [r11, r9]
subs r7, r7, #0x1
bne set_mmdc_io_lpm
cmp r3, #IMX_DDR_TYPE_LPDDR2
bne set_mmdc_io_lpm_done
ldr r6, =0x1000
ldr r9, [r8], #0x8
str r6, [r11, r9]
ldr r9, [r8], #0x8
str r6, [r11, r9]
ldr r6, =0x80000
ldr r9, [r8]
str r6, [r11, r9]
set_mmdc_io_lpm_done:
/*
* mask all GPC interrupts before
* enabling the RBC counters to
* avoid the counter starting too
* early if an interupt is already
* pending.
*/
ldr r11, [r0, #PM_INFO_MX6Q_GPC_V_OFFSET]
ldr r6, [r11, #MX6Q_GPC_IMR1]
ldr r7, [r11, #MX6Q_GPC_IMR2]
ldr r8, [r11, #MX6Q_GPC_IMR3]
ldr r9, [r11, #MX6Q_GPC_IMR4]
ldr r10, =0xffffffff
str r10, [r11, #MX6Q_GPC_IMR1]
str r10, [r11, #MX6Q_GPC_IMR2]
str r10, [r11, #MX6Q_GPC_IMR3]
str r10, [r11, #MX6Q_GPC_IMR4]
/*
* enable the RBC bypass counter here
* to hold off the interrupts. RBC counter
* = 32 (1ms), Minimum RBC delay should be
* 400us for the analog LDOs to power down.
*/
ldr r11, [r0, #PM_INFO_MX6Q_CCM_V_OFFSET]
ldr r10, [r11, #MX6Q_CCM_CCR]
bic r10, r10, #(0x3f << 21)
orr r10, r10, #(0x20 << 21)
str r10, [r11, #MX6Q_CCM_CCR]
/* enable the counter. */
ldr r10, [r11, #MX6Q_CCM_CCR]
orr r10, r10, #(0x1 << 27)
str r10, [r11, #MX6Q_CCM_CCR]
/* unmask all the GPC interrupts. */
ldr r11, [r0, #PM_INFO_MX6Q_GPC_V_OFFSET]
str r6, [r11, #MX6Q_GPC_IMR1]
str r7, [r11, #MX6Q_GPC_IMR2]
str r8, [r11, #MX6Q_GPC_IMR3]
str r9, [r11, #MX6Q_GPC_IMR4]
/*
* now delay for a short while (3usec)
* ARM is at 1GHz at this point
* so a short loop should be enough.
* this delay is required to ensure that
* the RBC counter can start counting in
* case an interrupt is already pending
* or in case an interrupt arrives just
* as ARM is about to assert DSM_request.
*/
ldr r6, =2000
rbc_loop:
subs r6, r6, #0x1
bne rbc_loop
/* Zzz, enter stop mode */
wfi
nop
nop
nop
nop
/*
* run to here means there is pending
* wakeup source, system should auto
* resume, we need to restore MMDC IO first
*/
mov r5, #0x0
resume_mmdc
/* return to suspend finish */
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
resume:
/* invalidate L1 I-cache first */
mov r6, #0x0
mcr p15, 0, r6, c7, c5, 0
mcr p15, 0, r6, c7, c5, 6
/* enable the Icache and branch prediction */
mov r6, #0x1800
mcr p15, 0, r6, c1, c0, 0
isb
/* get physical resume address from pm_info. */
ldr lr, [r0, #PM_INFO_RESUME_ADDR_OFFSET]
/* clear core0's entry and parameter */
ldr r11, [r0, #PM_INFO_MX6Q_SRC_P_OFFSET]
mov r7, #0x0
str r7, [r11, #MX6Q_SRC_GPR1]
str r7, [r11, #MX6Q_SRC_GPR2]
ldr r3, [r0, #PM_INFO_DDR_TYPE_OFFSET]
mov r5, #0x1
resume_mmdc
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(imx6_suspend)
/*
* The following code must assume it is running from physical address
* where absolute virtual addresses to the data section have to be
* turned into relative ones.
*/
ENTRY(v7_cpu_resume)
bl v7_invalidate_l1
#ifdef CONFIG_CACHE_L2X0
bl l2c310_early_resume
#endif
b cpu_resume
ENDPROC(v7_cpu_resume)