From patchwork Wed Aug 30 02:55:47 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Nicolas Pitre X-Patchwork-Id: 9928673 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id 6FE526032A for ; Wed, 30 Aug 2017 02:59:04 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 62FED21327 for ; Wed, 30 Aug 2017 02:59:04 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 57DE022B39; Wed, 30 Aug 2017 02:59:04 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-2.6 required=2.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,RCVD_IN_DNSWL_LOW autolearn=ham version=3.3.1 Received: from bombadil.infradead.org (bombadil.infradead.org [65.50.211.133]) (using TLSv1.2 with cipher AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.wl.linuxfoundation.org (Postfix) with ESMTPS id 71C5B21327 for ; Wed, 30 Aug 2017 02:59:03 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=lists.infradead.org; s=bombadil.20170209; h=Sender: Content-Transfer-Encoding:Content-Type:MIME-Version:Cc:List-Subscribe: List-Help:List-Post:List-Archive:List-Unsubscribe:List-Id:References: In-Reply-To:Message-Id:Date:Subject:To:From:Reply-To:Content-ID: Content-Description:Resent-Date:Resent-From:Resent-Sender:Resent-To:Resent-Cc :Resent-Message-ID:List-Owner; bh=YC+JoIlh1xeglGj1D1d6qF9vJ7yYMsIAewf8Ruygi90=; b=AD8eufSt6n+HtKAI373c8tTQTe zIhgQSfw9wHkXc3gW9sP/PoALZ8EdlJiYliyetks4jZArkLAk/C5eQnDXr2p7Xhp4LSOvhvGYmAlm tq16jSprzM/jg4M6LtF6K3N7lKEy47KllNiz3BozAUMtrJfQFLqymvYdOxZ3iZiunwm+NSAXxT6Fv hnZpkVoKXtrWOAXDHdGb4lE/EasbWnIjaXR/VJx+25mdMcV6ULWul9P5vXWQqatGm3BIO1K+JFbYU gR6n99KIdeS9r4lobAVPRhjR0O0r9iPspvnQ7QyGGTsLOSTkmhfT2WrpDpbv/zM0Yo7GKDxQDtB69 xJ5EhfVA==; Received: from localhost ([127.0.0.1] helo=bombadil.infradead.org) by bombadil.infradead.org with esmtp (Exim 4.87 #1 (Red Hat Linux)) id 1dmtDX-00027k-Sx; Wed, 30 Aug 2017 02:58:59 +0000 Received: from pb-smtp2.pobox.com ([64.147.108.71] helo=sasl.smtp.pobox.com) by bombadil.infradead.org with esmtps (Exim 4.87 #1 (Red Hat Linux)) id 1dmtAs-0008C9-OR for linux-arm-kernel@lists.infradead.org; Wed, 30 Aug 2017 02:56:22 +0000 Received: from sasl.smtp.pobox.com (unknown [127.0.0.1]) by pb-smtp2.pobox.com (Postfix) with ESMTP id 6B0C9A4B61; Tue, 29 Aug 2017 22:55:53 -0400 (EDT) DKIM-Signature: v=1; a=rsa-sha1; c=relaxed; d=pobox.com; h=from:to:cc :subject:date:message-id:in-reply-to:references; s=sasl; bh=4k3L 1Hmh7dztebZH4F1qje/kSMo=; b=GT6W9dCheW0HADI8TsF2TZoWXjjznnTtvc4W hxuiphkt852+L/BPMz0jZ8kDeLYrTTJGhYhvVqoVMnCuksceikeS+WZndvbUb1vg VXDGDp8SqNmvpTDJh8BLGJ1XutQOVJdgeNKcn8zxuMr2NMdFYOGCQYo8iK4ixfTE d8Chi/Q= Received: from pb-smtp2.nyi.icgroup.com (unknown [127.0.0.1]) by pb-smtp2.pobox.com (Postfix) with ESMTP id 63585A4B60; Tue, 29 Aug 2017 22:55:53 -0400 (EDT) Received: from yoda.home (unknown [70.80.200.199]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by pb-smtp2.pobox.com (Postfix) with ESMTPSA id E43A1A4B5F; Tue, 29 Aug 2017 22:55:52 -0400 (EDT) Received: from xanadu.home (xanadu.home [192.168.2.2]) by yoda.home (Postfix) with ESMTP id B739E2DA0696; Tue, 29 Aug 2017 22:55:51 -0400 (EDT) From: Nicolas Pitre To: linux-arm-kernel@lists.infradead.org Subject: [PATCH v2 5/5] ARM: XIP kernel: store .data compressed in ROM Date: Tue, 29 Aug 2017 22:55:47 -0400 Message-Id: <20170830025547.30347-6-nicolas.pitre@linaro.org> X-Mailer: git-send-email 2.9.5 In-Reply-To: <20170830025547.30347-1-nicolas.pitre@linaro.org> References: <20170830025547.30347-1-nicolas.pitre@linaro.org> X-Pobox-Relay-ID: BC8D5B42-8D2E-11E7-8284-9D2B0D78B957-78420484!pb-smtp2.pobox.com X-CRM114-Version: 20100106-BlameMichelson ( TRE 0.8.0 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20170829_195615_083651_93F808D1 X-CRM114-Status: GOOD ( 27.14 ) X-BeenThere: linux-arm-kernel@lists.infradead.org X-Mailman-Version: 2.1.21 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Chris Brandt , Russell King - ARM Linux , Arnd Bergmann , Ard Biesheuvel MIME-Version: 1.0 Sender: "linux-arm-kernel" Errors-To: linux-arm-kernel-bounces+patchwork-linux-arm=patchwork.kernel.org@lists.infradead.org X-Virus-Scanned: ClamAV using ClamSMTP The .data segment stored in ROM is only copied to RAM once at boot time and never referenced afterwards. This is arguably a suboptimal usage of ROM resources. This patch allows for compressing the .data segment before storing it into ROM and decompressing it to RAM rather than simply copying it, saving on precious ROM space. Because global data is not available yet (obviously) we must allocate decompressor workspace memory on the stack. The .bss area is used as a stack area for that purpose before it is cleared. The required stack frame is 9568 bytes for __inflate_kernel_data() alone, so make sure the .bss is large enough to cope with that plus extra room for called functions or fail the build. Those numbers were picked arbitrarily based on the above 9568 byte stack frame: 10240 (2.5 * PAGE_SIZE): used to override -Wframe-larger-than whose default value is 1024. 12288 (3 * PAGE_SIZE): minimum .bss size to contain the stack. Signed-off-by: Nicolas Pitre Reviewed-by: Ard Biesheuvel --- arch/arm/Kconfig | 11 +++++++ arch/arm/boot/Makefile | 13 +++++++- arch/arm/boot/deflate_xip_data.sh | 64 +++++++++++++++++++++++++++++++++++++ arch/arm/kernel/Makefile | 5 +++ arch/arm/kernel/head-common.S | 11 ++++++- arch/arm/kernel/head-inflate-data.c | 62 +++++++++++++++++++++++++++++++++++ arch/arm/kernel/vmlinux-xip.lds.S | 8 +++++ 7 files changed, 172 insertions(+), 2 deletions(-) create mode 100755 arch/arm/boot/deflate_xip_data.sh create mode 100644 arch/arm/kernel/head-inflate-data.c diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig index 61a0cb1506..bf79c461bd 100644 --- a/arch/arm/Kconfig +++ b/arch/arm/Kconfig @@ -2005,6 +2005,17 @@ config XIP_PHYS_ADDR be linked for and stored to. This address is dependent on your own flash usage. +config XIP_DEFLATED_DATA + bool "Store kernel .data section compressed in ROM" + depends on XIP_KERNEL + select ZLIB_INFLATE + help + Before the kernel is actually executed, its .data section has to be + copied to RAM from ROM. This option allows for storing that data + in compressed form and decompressed to RAM rather than merely being + copied, saving some precious ROM space. A possible drawback is a + slightly longer boot delay. + config KEXEC bool "Kexec system call (EXPERIMENTAL)" depends on (!SMP || PM_SLEEP_SMP) diff --git a/arch/arm/boot/Makefile b/arch/arm/boot/Makefile index 50f8d1be7f..a3af4dc08c 100644 --- a/arch/arm/boot/Makefile +++ b/arch/arm/boot/Makefile @@ -31,8 +31,19 @@ targets := Image zImage xipImage bootpImage uImage ifeq ($(CONFIG_XIP_KERNEL),y) +cmd_deflate_xip_data = $(CONFIG_SHELL) -c \ + '$(srctree)/$(src)/deflate_xip_data.sh $< $@ || { rm -f $@; false; }' + +ifeq ($(CONFIG_XIP_DEFLATED_DATA),y) +quiet_cmd_mkxip = XIPZ $@ +cmd_mkxip = $(cmd_objcopy) && $(cmd_deflate_xip_data) +else +quiet_cmd_mkxip = $(quiet_cmd_objcopy) +cmd_mkxip = $(cmd_objcopy) +endif + $(obj)/xipImage: vmlinux FORCE - $(call if_changed,objcopy) + $(call if_changed,mkxip) @$(kecho) ' Physical Address of xipImage: $(CONFIG_XIP_PHYS_ADDR)' $(obj)/Image $(obj)/zImage: FORCE diff --git a/arch/arm/boot/deflate_xip_data.sh b/arch/arm/boot/deflate_xip_data.sh new file mode 100755 index 0000000000..1189598a25 --- /dev/null +++ b/arch/arm/boot/deflate_xip_data.sh @@ -0,0 +1,64 @@ +#!/bin/sh + +# XIP kernel .data segment compressor +# +# Created by: Nicolas Pitre, August 2017 +# Copyright: (C) 2017 Linaro Limited +# +# 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. + +# This script locates the start of the .data section in xipImage and +# substitutes it with a compressed version. The needed offsets are obtained +# from symbol addresses in vmlinux. It is expected that .data extends to +# the end of xipImage. + +set -e + +VMLINUX="$1" +XIPIMAGE="$2" + +DD="dd status=none" + +# Use "make V=1" to debug this script. +case "$KBUILD_VERBOSE" in +*1*) + set -x + ;; +esac + +sym_val() { + # extract hex value for symbol in $1 + local val=$($NM "$VMLINUX" | sed -n "/ $1$/{s/ .*$//p;q}") + [ "$val" ] || { echo "can't find $1 in $VMLINUX" 1>&2; exit 1; } + # convert from hex to decimal + echo $((0x$val)) +} + +__data_loc=$(sym_val __data_loc) +_edata_loc=$(sym_val _edata_loc) +base_offset=$(sym_val _xiprom) + +# convert to file based offsets +data_start=$(($__data_loc - $base_offset)) +data_end=$(($_edata_loc - $base_offset)) + +# Make sure data occupies the last part of the file. +file_end=$(stat -c "%s" "$XIPIMAGE") +if [ "$file_end" != "$data_end" ]; then + printf "end of xipImage doesn't match with _edata_loc (%#x vs %#x)\n" \ + $(($file_end + $base_offset)) $_edata_loc 2>&1 + exit 1; +fi + +# be ready to clean up +trap 'rm -f "$XIPIMAGE.tmp"' 0 1 2 3 + +# substitute the data section by a compressed version +$DD if="$XIPIMAGE" count=$data_start iflag=count_bytes of="$XIPIMAGE.tmp" +$DD if="$XIPIMAGE" skip=$data_start iflag=skip_bytes | +gzip -9 >> "$XIPIMAGE.tmp" + +# replace kernel binary +mv -f "$XIPIMAGE.tmp" "$XIPIMAGE" diff --git a/arch/arm/kernel/Makefile b/arch/arm/kernel/Makefile index ad325a8c7e..52f437997c 100644 --- a/arch/arm/kernel/Makefile +++ b/arch/arm/kernel/Makefile @@ -87,6 +87,11 @@ head-y := head$(MMUEXT).o obj-$(CONFIG_DEBUG_LL) += debug.o obj-$(CONFIG_EARLY_PRINTK) += early_printk.o +# This is executed very early using a temporary stack when no memory allocator +# nor global data is available. Everything has to be allocated on the stack. +CFLAGS_head-inflate-data.o := $(call cc-option,-Wframe-larger-than=10240) +obj-$(CONFIG_XIP_DEFLATED_DATA) += head-inflate-data.o + obj-$(CONFIG_ARM_VIRT_EXT) += hyp-stub.o AFLAGS_hyp-stub.o :=-Wa,-march=armv7-a ifeq ($(CONFIG_ARM_PSCI),y) diff --git a/arch/arm/kernel/head-common.S b/arch/arm/kernel/head-common.S index bf9c4e38ec..a25027b87a 100644 --- a/arch/arm/kernel/head-common.S +++ b/arch/arm/kernel/head-common.S @@ -87,7 +87,14 @@ __mmap_switched: adr r4, __mmap_switched_data mov fp, #0 -#ifdef CONFIG_XIP_KERNEL +#if defined(CONFIG_XIP_DEFLATED_DATA) + ARM( ldr sp, [r4], #4 ) + THUMB( ldr sp, [r4] ) + THUMB( add r4, #4 ) + bl __inflate_kernel_data @ decompress .data to RAM + teq r0, #0 + bne __error +#elif defined(CONFIG_XIP_KERNEL) ARM( ldmia r4!, {r0, r1, r2, sp} ) THUMB( ldmia r4!, {r0, r1, r2, r3} ) THUMB( mov sp, r3 ) @@ -114,9 +121,11 @@ ENDPROC(__mmap_switched) .type __mmap_switched_data, %object __mmap_switched_data: #ifdef CONFIG_XIP_KERNEL +#ifndef CONFIG_XIP_DEFLATED_DATA .long _sdata @ r0 .long __data_loc @ r1 .long _edata_loc @ r2 +#endif .long __bss_stop @ sp (temporary stack in .bss) #endif diff --git a/arch/arm/kernel/head-inflate-data.c b/arch/arm/kernel/head-inflate-data.c new file mode 100644 index 0000000000..4598b959c9 --- /dev/null +++ b/arch/arm/kernel/head-inflate-data.c @@ -0,0 +1,62 @@ +/* + * XIP kernel .data segment decompressor + * + * Created by: Nicolas Pitre, August 2017 + * Copyright: (C) 2017 Linaro Limited + * + * 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. + */ + +#include +#include + +/* for struct inflate_state */ +#include "../../../lib/zlib_inflate/inftrees.h" +#include "../../../lib/zlib_inflate/inflate.h" +#include "../../../lib/zlib_inflate/infutil.h" + +extern char __data_loc[]; +extern char _edata_loc[]; +extern char _sdata[]; + +/* + * This code is called very early during the boot process to decompress + * the .data segment stored compressed in ROM. Therefore none of the global + * variables are valid yet, hence no kernel services such as memory + * allocation is available. Everything must be allocated on the stack and + * we must avoid any global data access. We use a temporary stack located + * in the .bss area. The linker script makes sure the .bss is big enough + * to hold our stack frame plus some room for called functions. + * + * We mimic the code in lib/decompress_inflate.c to use the smallest work + * area possible. And because everything is statically allocated on the + * stack then there is no need to clean up before returning. + */ + +int __init __inflate_kernel_data(void) +{ + struct z_stream_s stream, *strm = &stream; + struct inflate_state state; + char *in = __data_loc; + int rc; + + /* Check and skip gzip header (assume no filename) */ + if (in[0] != 0x1f || in[1] != 0x8b || in[2] != 0x08 || in[3] & ~3) + return -1; + in += 10; + + strm->workspace = &state; + strm->next_in = in; + strm->avail_in = _edata_loc - __data_loc; /* upper bound */ + strm->next_out = _sdata; + strm->avail_out = _edata_loc - __data_loc; + zlib_inflateInit2(strm, -MAX_WBITS); + WS(strm)->inflate_state.wsize = 0; + WS(strm)->inflate_state.window = NULL; + rc = zlib_inflate(strm, Z_FINISH); + if (rc == Z_STREAM_END) + rc = strm->avail_out; /* should be 0 */ + return rc; +} diff --git a/arch/arm/kernel/vmlinux-xip.lds.S b/arch/arm/kernel/vmlinux-xip.lds.S index 39b1fb470a..7a84431008 100644 --- a/arch/arm/kernel/vmlinux-xip.lds.S +++ b/arch/arm/kernel/vmlinux-xip.lds.S @@ -306,3 +306,11 @@ ASSERT((__arch_info_end - __arch_info_begin), "no machine record defined") */ ASSERT(__hyp_idmap_text_end - (__hyp_idmap_text_start & PAGE_MASK) <= PAGE_SIZE, "HYP init code too big or misaligned") + +#ifdef CONFIG_XIP_DEFLATED_DATA +/* + * The .bss is used as a stack area for __inflate_kernel_data() whose stack + * frame is 9568 bytes. Make sure it has extra room left. + */ +ASSERT((_end - __bss_start) >= 12288, ".bss too small for CONFIG_XIP_DEFLATED_DATA") +#endif