From patchwork Thu May 19 17:31:28 2011 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Frank Hofmann X-Patchwork-Id: 798222 Received: from bombadil.infradead.org (bombadil.infradead.org [18.85.46.34]) by demeter2.kernel.org (8.14.4/8.14.3) with ESMTP id p4JHXxHJ021816 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO) for ; Thu, 19 May 2011 17:34:21 GMT Received: from canuck.infradead.org ([2001:4978:20e::1]) by bombadil.infradead.org with esmtps (Exim 4.76 #1 (Red Hat Linux)) id 1QN74U-00081s-JL; Thu, 19 May 2011 17:31:38 +0000 Received: from localhost ([127.0.0.1] helo=canuck.infradead.org) by canuck.infradead.org with esmtp (Exim 4.76 #1 (Red Hat Linux)) id 1QN74S-00076k-Qp; Thu, 19 May 2011 17:31:36 +0000 Received: from mxvs2.esa.t-systems.com ([81.7.202.143]) by canuck.infradead.org with esmtp (Exim 4.76 #1 (Red Hat Linux)) id 1QN74N-00076G-6y for linux-arm-kernel@lists.infradead.org; Thu, 19 May 2011 17:31:34 +0000 Received: from unknown (HELO nl-exc-01.intra.local) ([82.210.235.24]) by mx.esa.t-systems.com with ESMTP; 19 May 2011 17:31:29 +0000 Received: from magrathea ([10.101.8.37]) by nl-exc-01.intra.local with Microsoft SMTPSVC(6.0.3790.3959); Thu, 19 May 2011 19:31:28 +0200 Date: Thu, 19 May 2011 18:31:28 +0100 (BST) From: Frank Hofmann To: linux-arm-kernel@lists.infradead.org, linux-pm@lists.linux-foundation.org, tuxonice-devel@tuxonice.net Subject: [RFC PATCH] ARM hibernation / suspend-to-disk support code In-Reply-To: <3DCE2F529B282E4B8F53D4D8AA406A07014FFE@008-AM1MPN1-022.mgdnok.nokia.com> Message-ID: References: <3DCE2F529B282E4B8F53D4D8AA406A07014FFE@008-AM1MPN1-022.mgdnok.nokia.com> User-Agent: Alpine 2.00 (DEB 1167 2008-08-23) MIME-Version: 1.0 X-OriginalArrivalTime: 19 May 2011 17:31:28.0863 (UTC) FILETIME=[961AC6F0:01CC164A] X-CRM114-Version: 20090807-BlameThorstenAndJenny ( TRE 0.7.6 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20110519_133131_806477_3F7B2EAF X-CRM114-Status: GOOD ( 34.89 ) X-Spam-Score: -0.7 (/) X-Spam-Report: SpamAssassin version 3.3.1 on canuck.infradead.org summary: Content analysis details: (-0.7 points) pts rule name description ---- ---------------------- -------------------------------------------------- -0.7 RCVD_IN_DNSWL_LOW RBL: Sender listed at http://www.dnswl.org/, low trust [81.7.202.143 listed in list.dnswl.org] 0.0 NORMAL_HTTP_TO_IP URI: Uses a dotted-decimal IP address in URL X-BeenThere: linux-arm-kernel@lists.infradead.org X-Mailman-Version: 2.1.12 Precedence: list Reply-To: frank.hofmann@tomtom.com List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: linux-arm-kernel-bounces@lists.infradead.org Errors-To: linux-arm-kernel-bounces+patchwork-linux-arm=patchwork.kernel.org@lists.infradead.org X-Greylist: IP, sender and recipient auto-whitelisted, not delayed by milter-greylist-4.2.6 (demeter2.kernel.org [140.211.167.43]); Thu, 19 May 2011 17:34:21 +0000 (UTC) Hi, /me again ... Sorry that this took a little ... holidays. And work. And distractions... Anyway, here we go again, basic code to enable hibernation (suspend-to-disk) on ARM platforms. Any comments highly welcome. To use this, you need sleep.S modifications for your SoC type (to get __save/__restore_processor_state hooks). I've sent some of those for illustration earlier, they haven't changed, I've not included them here, so pick these changes up from: http://68.183.106.108/lists/linux-pm/msg24020.html The patch below only contains the _generic_ code. This is tested on S5P6450 and OMAP3, with the sleep...S changes just mentioned - check the archives for those. Works both with normal swsusp and tuxonice (again, check the archives for the TOI modification needed). Previously, I've reported OMAP3 video issues, after resume-from-disk. That isn't fully solved (it's a driver issue) but I've found a workaround: Trigger the resume from initramfs, after loading a logo image into the framebuffer and switching it on. That gets everything back without corruptions / wrong LCD reinitialization. The OMAP video seems a bit of a diva; I've got one board type on which suspend/resume work perfectly but the omapdss driver spits out thousands of error interrupts during system startup (before the image is loaded), and the other board where all that is fine but the restore somehow garbles the LCD clocking (but the driver's sysfs files claim it's the same). In short: This stuff really works now, for all I can say. And adding support for new type of ARM SoC doesn't touch the basic / generic code at all anymore either. Anyway ... About the patch, changes vs. all previous suggestions: * Made the assembly sources as small as I responsibly could ;-) They compile for thumb2 (but I haven't tested that yet) as well. * The page copy loop is now a C function. That also has the advantage that one can use cpu_switch_mm() - a macro - there for swapper_pg_dir, which makes resume via uswsusp ioctl or /sys/power/tuxonice/do_resume possible (only tested the latter, though). * The SoC state save/restore is made to (re-)use the existing code in sleep....S for the particular chip. OMAP3 and S5P64xx are provided as examples of that. * The save/restore_processor_state() hooks are now used in the same way as e.g. existing powerpc code uses them (to trigger lazy saves before and perform cache flushes after). Things that probably aren't perfect yet: * The code currently reserves a full page for the saved "core" state. This is more than absolutely necessary; anyone think it's a problem ? * it sets aside another half a page of __nosavedata page for use as temporary stack during the image copy (so that funcs can be called). Right now on ARM, that's not an issue because even with TuxOnIce in, there's less than 20 bytes of nosave stuff, so can as well put the rest of that page to good use ;-) * I'd love to get rid of the include/asm-generic/vmlinux.lds.h change, as it seems that's not necessary in other architectures. Without that, the code gives a link error when building vmlinux though, and I'm unsure how to address that. * The "integration" with the CPU sleep code is rather "backdoorish". While the hooks into ..._cpu_suspend aren't massive, and there's no code duplication, it'd be nicer to eventually have a cleaner way. * An OMAPDSS restore troubleshooting HOWTO would be helpful ;-) * The patch needs to be rebaselined against a current kernel; any preferences which tree to base this on ? Thanks for all help with the little nits ! FrankH. arch/arm/Kconfig | 3 + arch/arm/include/asm/memory.h | 1 + arch/arm/include/asm/suspend.h | 6 ++ arch/arm/kernel/cpu.c | 65 ++++++++++++++++++++++++++ arch/arm/kernel/swsusp.S | 92 +++++++++++++++++++++++++++++++++++++ arch/arm/kernel/vmlinux.lds.S | 3 +- include/asm-generic/vmlinux.lds.h | 2 +- 7 files changed, 170 insertions(+), 2 deletions(-) diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig index 6b6786c..859dd86 100644 --- a/arch/arm/Kconfig +++ b/arch/arm/Kconfig @@ -198,6 +198,9 @@ config VECTORS_BASE config ARCH_HAS_CPU_IDLE_WAIT def_bool y +config ARCH_HIBERNATION_POSSIBLE + def_bool n + source "init/Kconfig" source "kernel/Kconfig.freezer" diff --git a/arch/arm/include/asm/memory.h b/arch/arm/include/asm/memory.h index 5421d82..23e93a6 100644 --- a/arch/arm/include/asm/memory.h +++ b/arch/arm/include/asm/memory.h @@ -191,6 +191,7 @@ static inline void *phys_to_virt(unsigned long x) */ #define __pa(x) __virt_to_phys((unsigned long)(x)) #define __va(x) ((void *)__phys_to_virt((unsigned long)(x))) +#define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x),0)) #define pfn_to_kaddr(pfn) __va((pfn) << PAGE_SHIFT) /* diff --git a/arch/arm/include/asm/suspend.h b/arch/arm/include/asm/suspend.h new file mode 100644 index 0000000..7ab1fd2 --- /dev/null +++ b/arch/arm/include/asm/suspend.h @@ -0,0 +1,6 @@ +#ifndef __ASM_ARM_SUSPEND_H +#define __ASM_ARM_SUSPEND_H + +static inline int arch_prepare_suspend(void) { return 0; } + +#endif /* __ASM_ARM_SUSPEND_H */ diff --git a/arch/arm/kernel/cpu.c b/arch/arm/kernel/cpu.c new file mode 100644 index 0000000..764c8fa --- /dev/null +++ b/arch/arm/kernel/cpu.c @@ -0,0 +1,65 @@ +/* + * Hibernation support specific for ARM + * + * Derived from work on ARM hibernation support by: + * + * Ubuntu project, hibernation support for mach-dove + * Copyright (C) 2010 Nokia Corporation (Hiroshi Doyu) + * Copyright (C) 2010 Texas Instruments, Inc. (Teerth Reddy et al.) + * https://lkml.org/lkml/2010/6/18/4 + * https://lists.linux-foundation.org/pipermail/linux-pm/2010-June/027422.html + * https://patchwork.kernel.org/patch/96442/ + * + * Copyright (C) 2006 Rafael J. Wysocki + * + * License terms: GNU General Public License (GPL) version 2 + */ + +#include +#include +#include +#include + +extern const void __nosave_begin, __nosave_end; + +int pfn_is_nosave(unsigned long pfn) +{ + unsigned long nosave_begin_pfn = __pa_symbol(&__nosave_begin) >> PAGE_SHIFT; + unsigned long nosave_end_pfn = PAGE_ALIGN(__pa_symbol(&__nosave_end)) >> PAGE_SHIFT; + + return (pfn >= nosave_begin_pfn) && (pfn < nosave_end_pfn); +} + +void save_processor_state(void) +{ + flush_thread(); +} + +void restore_processor_state(void) +{ + local_flush_tlb_all(); +} + +u8 __swsusp_arch_ctx[PAGE_SIZE] __attribute__((aligned(PAGE_SIZE))); +u8 __swsusp_resume_stk[PAGE_SIZE/2] __nosavedata; + +/* + * The framework loads the hibernation image into this linked list, + * for swsusp_arch_resume() to copy back to the proper destinations. + * + * To make this work if resume is triggered from initramfs, the + * pagetables need to be switched to allow writes to kernel mem. + */ +void notrace __swsusp_arch_restore_prepare(void) +{ + cpu_switch_mm(__virt_to_phys(swapper_pg_dir), current->active_mm); +} + +void notrace __swsusp_arch_restore_image(void) +{ + extern struct pbe *restore_pblist; + struct pbe *pbe; + + for (pbe = restore_pblist; pbe; pbe = pbe->next) + copy_page(pbe->orig_address, pbe->address); +} diff --git a/arch/arm/kernel/swsusp.S b/arch/arm/kernel/swsusp.S new file mode 100644 index 0000000..fb260a7 --- /dev/null +++ b/arch/arm/kernel/swsusp.S @@ -0,0 +1,92 @@ +/* + * Hibernation support specific for ARM + * + * Based on work by: + * + * Ubuntu project, hibernation support for mach-dove, + * Copyright (C) 2010 Nokia Corporation (Hiroshi Doyu) + * Copyright (C) 2010 Texas Instruments, Inc. (Teerth Reddy et al.) + * https://lkml.org/lkml/2010/6/18/4 + * https://lists.linux-foundation.org/pipermail/linux-pm/2010-June/027422.html + * https://patchwork.kernel.org/patch/96442/ + * + * Copyright (C) 2006 Rafael J. Wysocki + * + * License terms: GNU General Public License (GPL) version 2 + */ + +#include +#include +#include +#include +#include + +/* + * Save the current CPU state before suspend / poweroff. + */ +ENTRY(swsusp_arch_suspend) + ldr r0, =__swsusp_arch_ctx + mrs r1, cpsr + str r1, [r0], #4 /* CPSR */ +ARM( msr cpsr_c, #SYSTEM_MODE ) +THUMB( mov r2, #SYSTEM_MODE ) +THUMB( msr cpsr_c, r2 ) + stm r0!, {r4-r12,lr} /* nonvolatile regs */ + str sp, [r0], #4 +ARM( msr cpsr_c, #SVC_MODE ) +THUMB( mov r2, #SVC_MODE ) +THUMB( msr cpsr_c, r2 ) + mrs r2, spsr + stm r0!, {r2,lr} /* SVC SPSR, SVC regs */ + str sp, [r0], #4 + msr cpsr, r1 /* restore mode at entry */ + push {lr} + bl __save_processor_state + pop {lr} + b swsusp_save +ENDPROC(swsusp_arch_suspend) + +/* + * Restore the memory image from the pagelists, and load the CPU registers + * from saved state. + */ +ENTRY(swsusp_arch_resume) + bl __swsusp_arch_restore_prepare + /* + * Switch stack to a nosavedata region to make sure image restore + * doesn't clobber it underneath itself. + */ + ldr sp, =(__swsusp_resume_stk + PAGE_SIZE / 2) + bl __swsusp_arch_restore_image + + /* + * Restore the CPU registers. + */ +ARM( msr cpsr_c, #SYSTEM_MODE ) +THUMB( mov r2, #SYSTEM_MODE ) +THUMB( msr cpsr_c, r2 ) + ldr r0, =__swsusp_arch_ctx + ldr r1, [r0], #4 + msr cpsr_cxsf, r1 + ldm r0!, {r4-r12,lr} + ldr sp, [r0], #4 +ARM( msr cpsr_c, #SVC_MODE ) +THUMB( mov r2, #SVC_MODE ) +THUMB( msr cpsr_c, r2 ) + ldm r0!, {r2,lr} + ldr sp, [r0], #4 + msr spsr_cxsf, r2 + msr cpsr_c, r1 + + /* + * From here on we have a valid stack again. Core state is + * not restored yet, redirect to the machine-specific + * implementation to get that done. + * Resume has succeeded at this point; if the machine-specific + * code wants to fail it needs to panic. + */ + mov r1, #0 + push {r1,lr} + bl __restore_processor_state /* restore core state */ + pop {r0,pc} +ENDPROC(swsusp_arch_resume) diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S index 4957e13..e691c77 100644 --- a/arch/arm/kernel/vmlinux.lds.S +++ b/arch/arm/kernel/vmlinux.lds.S @@ -153,7 +153,6 @@ SECTIONS __init_end = .; #endif - NOSAVE_DATA CACHELINE_ALIGNED_DATA(32) /* @@ -176,6 +175,8 @@ SECTIONS } _edata_loc = __data_loc + SIZEOF(.data); + NOSAVE_DATA + #ifdef CONFIG_HAVE_TCM /* * We align everything to a page boundary so we can diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index b6e818f..0d39ae0 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -171,7 +171,7 @@ #define NOSAVE_DATA \ . = ALIGN(PAGE_SIZE); \ VMLINUX_SYMBOL(__nosave_begin) = .; \ - *(.data.nosave) \ + .data.nosave : { *(.data.nosave) } \ . = ALIGN(PAGE_SIZE); \ VMLINUX_SYMBOL(__nosave_end) = .;