kexec: put bzImage and ramdisk above 4G for x86 64bit
Subsystems (2)
v5 cover kexec: put bzImage and ramdisk above 4G for x86 64bit
Now we have limit kdump reserved under 896M, because kexec has the limitation. and also bzImage need to stay under 4g.
kernel parts changes could be found at:
git://git.kernel.org/pub/scm/linux/kernel/git/yinghai/linux-yinghai.git for-x86-boot
here patches are for kexec tools to load bzImage and ramdisk above 4G acccording to new added boot header fields.
-v3: address review from Eric to use locate_hole at first.
use xloadflags instead.
-v4: remove the restriction about bzImage not crossing GB boundary.
add real-mode fix for bzImage.
add --entry-32bit and --real-mode for skip bzImage64.
-v5: use USE_EXT_BOOT_PARAMS bit in xloadflags.
Yinghai Lu (5):
kexec, x86: add boot header member for version 2.12 kexec, x86: clean boot_params area for entry-32bit path kexec, x86: Fix bzImage real-mode booting
kexec, x86: put ramdisk/cmd_line above 4G for 64bit bzImage kexec, x86_64: Load bzImage64 above 4G
include/x86/x86-linux.h | 32 +++- kexec/arch/i386/include/arch/options.h | 4 +- kexec/arch/i386/kexec-bzImage.c | 80 +++++++-- kexec/arch/i386/x86-linux-setup.c | 26 ++- kexec/arch/x86_64/Makefile | 1 + kexec/arch/x86_64/kexec-bzImage64.c | 335 ++++++++++++++++++++++++++++++++ kexec/arch/x86_64/kexec-x86_64.c | 1 + kexec/arch/x86_64/kexec-x86_64.h | 5 + 8 files changed, 462 insertions(+), 22 deletions(-) create mode 100644 kexec/arch/x86_64/kexec-bzImage64.c
Signature · 7 lines
1.7.7 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v5 1/5 kexec, x86: add boot header member for version 2.12
will use ext_ramdisk_image/size, and xloadflags to put ramdisk and bzImage high for 64bit.
--- include/x86/x86-linux.h | 32 +++++++++++++++++++++++++++----- 1 files changed, 27 insertions(+), 5 deletions(-)
include/x86/x86-linux.h +27 -5
@@ -32,7 +32,7 @@ struct drive_info_struct { };struct sys_desc_table {uint16_t length; -uint8_t table[318 ]; +uint8_t table[30 ]; };struct apm_bios_info {
@@ -112,6 +112,10 @@ struct x86_linux_param_header {struct apm_bios_info apm_bios_info;/* 0x40 */ struct drive_info_struct drive_info;/* 0x80 */ struct sys_desc_table sys_desc_table;/* 0xa0 */ +uint32_t ext_ramdisk_image;/* 0xc0 */ +uint32_t ext_ramdisk_size;/* 0xc4 */ +uint32_t ext_cmd_line_ptr;/* 0xc8 */ +uint8_t reserved4_1[0x1e0 -0xcc ];/* 0xcc */ uint32_t alt_mem_k;/* 0x1e0 */ uint8_t reserved5[4 ];/* 0x1e4 */ uint8_t e820_map_nr;/* 0x1e8 */
@@ -174,11 +178,18 @@ struct x86_linux_param_header {/* 2.04+ */ uint32_t kernel_alignment;/* 0x230 */ uint8_t relocatable_kernel;/* 0x234 */ -uint8_t reserved15[3 ];/* 0x235 */ +uint8_t min_alignment;/* 0x235 */ +uint16_t xloadflags;/* 0x236 */ uint32_t cmdline_size;/* 0x238 */ uint32_t hardware_subarch;/* 0x23C */ uint64_t hardware_subarch_data;/* 0x240 */ -uint8_t reserved16[0x290 -0x248 ];/* 0x248 */ +uint32_t payload_offset;/* 0x248 */ +uint32_t payload_length;/* 0x24C */ +uint64_t setup_data;/* 0x250 */ +uint64_t pref_address;/* 0x258 */ +uint32_t init_size;/* 0x260 */ +uint32_t handover_offset;/* 0x264 */ +uint8_t reserved16[0x290 -0x268 ];/* 0x268 */ uint32_t edd_mbr_sig_buffer[EDD_MBR_SIG_MAX ];/* 0x290 */ #endif struct e820entry e820_map[E820MAX ];/* 0x2d0 */
@@ -195,7 +206,11 @@ struct x86_linux_faked_param_header { };struct x86_linux_header { -uint8_t reserved1[0x1f1 ];/* 0x000 */ +uint8_t reserved1[0xc0 ];/* 0x000 */ +uint32_t ext_ramdisk_image;/* 0x0c0 */ +uint32_t ext_ramdisk_size;/* 0x0c4 */ +uint32_t ext_cmd_line_ptr;/* 0x0c8 */ +uint8_t reserved1_1[0x1f1 -0xcc ];/* 0x0cc */ uint8_t setup_sects;/* 0x1f1 */ uint16_t root_flags;/* 0x1f2 */ uint16_t syssize;/* 0x1f4 */
@@ -241,10 +256,17 @@ struct x86_linux_header {#else uint32_t kernel_alignment;/* 0x230 */ uint8_t relocatable_kernel;/* 0x234 */ -uint8_t reserved6[3 ];/* 0x235 */ +uint8_t min_alignment;/* 0x235 */ +uint16_t xloadflags;/* 0x236 */ uint32_t cmdline_size;/* 0x238 */ uint32_t hardware_subarch;/* 0x23C */ uint64_t hardware_subarch_data;/* 0x240 */ +uint32_t payload_offset;/* 0x248 */ +uint32_t payload_length;/* 0x24C */ +uint64_t setup_data;/* 0x250 */ +uint64_t pref_address;/* 0x258 */ +uint32_t init_size;/* 0x260 */ +uint32_t handover_offset;/* 0x264 */ #endif }PACKED ;
Signature · 7 lines
1.7.7 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v5 2/5 kexec, x86: clean boot_params area for entry-32bit path
In reply to Yinghai Lukexec bzImage path setup data is shared with real-mode path, and setup_header is copied together with setup_code.
Later 32bit just use whole area as boot_params for real_mode_data. but those area for boot_params around setup_header is not cleaned that will leave some field in boot_param as non-zero value.
So clean around setup_header area for non real-mode entry path.
--- kexec/arch/i386/kexec-bzImage.c | 17 +++++++++++++++++ 1 files changed, 17 insertions(+), 0 deletions(-)
kexec/arch/i386/kexec-bzImage.c +17 -0
@@ -98,6 +98,19 @@ void bzImage_usage(void) } +static void clean_boot_params (unsigned char *real_mode,unsigned long size) +{ +unsigned long end; + +/* clear value before header */ +memset (real_mode,0 ,0x1f1 ); +/* clear value after setup_header */ + end = *(real_mode +0x201 ); + end +=0x202 ; +if (end < size) +memset (real_mode + end,0 , size - end); +} + int do_bzImage_load(struct kexec_info *info, const char *kernel, off_t kernel_len, const char *command_line, off_t command_line_len,
@@ -212,6 +225,10 @@ int do_bzImage_load(struct kexec_info *info, setup_size = kern16_size + command_line_len +PURGATORY_CMDLINE_SIZE ; real_mode =xmalloc (setup_size);memcpy (real_mode, kernel, kern16_size); +if (!real_mode_entry) { +clean_boot_params ((unsigned char *)real_mode, kern16_size); + real_mode->xloadflags |= (1 <<15 );/* USE_EXT_BOOT_PARAMS */ + }if (info->kexec_flags & (KEXEC_ON_CRASH |KEXEC_PRESERVE_CONTEXT )) {/* If using bzImage for capture kernel, then we will not be
Signature · 7 lines
1.7.7 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v5 3/5 kexec, x86: Fix bzImage real-mode booting
In reply to Yinghai LuWe need to keep space for bss, heap/stack before command line. otherwise command_line will be cleared by kernel 16bit init code.
also need to set 32bit start in real_mode header, kernel 16bit code need to jump there.
Also don't touch regs16 if --real-mode is not specified.
--- kexec/arch/i386/kexec-bzImage.c | 63 +++++++++++++++++++++++++++++++------- 1 files changed, 51 insertions(+), 12 deletions(-)
kexec/arch/i386/kexec-bzImage.c +51 -12
@@ -130,6 +130,8 @@ int do_bzImage_load(struct kexec_info *info,unsigned long kernel32_load_addr;char *modified_cmdline;unsigned long cmdline_end; +unsigned long kern16_size_needed; +unsigned long heap_size =0 ;/* * Find out about the file I am about to load.
@@ -221,9 +223,31 @@ int do_bzImage_load(struct kexec_info *info,elf_rel_build_load (info, &info->rhdr, purgatory, purgatory_size,0x3000 ,640 *1024 ,-1 ,0 );dbgprintf ("Loaded purgatory at addr 0x%lx\n" , info->rhdr.rel_addr); +/* The argument/parameter segment */ - setup_size = kern16_size + command_line_len +PURGATORY_CMDLINE_SIZE ; +if (real_mode_entry) { +/* need to include size for bss and heap etc */ +if (setup_header.protocol_version >=0x0201 ) + kern16_size_needed = setup_header.heap_end_ptr; +else + kern16_size_needed = kern16_size +8192 ;/* bss */ +if (kern16_size_needed < kern16_size) + kern16_size_needed = kern16_size; +if (kern16_size_needed >0xfffc ) +die ("kern16_size_needed is more then 64k\n" ); + heap_size =0xfffc - kern16_size_needed;/* less 64k */ + heap_size &= ~(0x200 -1 ); + kern16_size_needed += heap_size; + }else { + kern16_size_needed = kern16_size; +/* need to bigger than size of struct bootparams */ +if (kern16_size_needed <4096 ) + kern16_size_needed =4096 ; + } + setup_size = kern16_size_needed + command_line_len + +PURGATORY_CMDLINE_SIZE ; real_mode =xmalloc (setup_size); +memset (real_mode,0 , setup_size);memcpy (real_mode, kernel, kern16_size);if (!real_mode_entry) {clean_boot_params ((unsigned char *)real_mode, kern16_size);
@@ -282,11 +306,20 @@ int do_bzImage_load(struct kexec_info *info,/* Tell the kernel what is going on */ setup_linux_bootloader_parameters (info, real_mode, setup_base, - kern16_size, command_line, command_line_len, + kern16_size_needed, command_line, command_line_len, initrd, initrd_len); +if (real_mode_entry) { +/* restore can use heap and load high */ + real_mode->loader_flags =0x81 ; +if (real_mode->protocol_version >=0x0201 ) + real_mode->heap_end_ptr += heap_size -0x200 ;/*stack*/ + } +/* Get the initial register values */ -elf_rel_get_symbol (&info->rhdr,"entry16_regs" , ®s16,sizeof (regs16)); +if (real_mode_entry) +elf_rel_get_symbol (&info->rhdr,"entry16_regs" , + ®s16,sizeof (regs16));elf_rel_get_symbol (&info->rhdr,"entry32_regs" , ®s32,sizeof (regs32));/*
@@ -305,16 +338,18 @@ int do_bzImage_load(struct kexec_info *info,/* * Initialize the 16bit start information. */ - regs16.ds = regs16.es = regs16.fs = regs16.gs = setup_base >>4 ; - regs16.cs = regs16.ds +0x20 ; - regs16.ip =0 ; -/* XXX: Documentation/i386/boot.txt says 'ss' must equal 'ds' */ - regs16.ss = (elf_rel_get_addr (&info->rhdr,"stack_end" ) -64 *1024 ) >>4 ; -/* XXX: Documentation/i386/boot.txt says 'sp' must equal heap_end */ - regs16.esp =0xFFFC ;if (real_mode_entry) { + regs16.ds = regs16.es = regs16.fs = regs16.gs = setup_base >>4 ; + regs16.cs = regs16.ds +0x20 ; + regs16.ip =0 ; +/* XXX: Documentation/i386/boot.txt says 'ss' must equal 'ds' */ + regs16.ss = (elf_rel_get_addr (&info->rhdr,"stack_end" ) -64 *1024 ) >>4 ; +/* XXX: Documentation/i386/boot.txt says 'sp' must equal heap_end */ + regs16.esp =0xFFFC ; +printf ("Starting the kernel in real mode\n" ); regs32.eip =elf_rel_get_addr (&info->rhdr,"entry16" ); + real_mode->kernel_start = kernel32_load_addr; }if (real_mode_entry && kexec_debug) {unsigned long entry16_debug, pre32, first32;
@@ -334,8 +369,12 @@ int do_bzImage_load(struct kexec_info *info, regs32.eip = entry16_debug; } -elf_rel_set_symbol (&info->rhdr,"entry16_regs" , ®s16,sizeof (regs16)); -elf_rel_set_symbol (&info->rhdr,"entry16_debug_regs" , ®s16,sizeof (regs16)); +if (real_mode_entry) { +elf_rel_set_symbol (&info->rhdr,"entry16_regs" , + ®s16,sizeof (regs16)); +elf_rel_set_symbol (&info->rhdr,"entry16_debug_regs" , + ®s16,sizeof (regs16)); + }elf_rel_set_symbol (&info->rhdr,"entry32_regs" , ®s32,sizeof (regs32)); cmdline_end = setup_base + kern16_size + command_line_len -1 ; elf_rel_set_symbol(&info->rhdr, "cmdline_end", &cmdline_end,
Signature · 7 lines
1.7.7 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v5 4/5 kexec, x86: put ramdisk/cmd_line above 4G for 64bit bzImage
In reply to Yinghai LuWe could put ramdisk/cmdline above for bzImage on 64bit for protocol 2.12.
-v2: change ext_... handling to way that eric like.
--- kexec/arch/i386/x86-linux-setup.c | 26 ++++++++++++++++++++++---- 1 files changed, 22 insertions(+), 4 deletions(-)
kexec/arch/i386/x86-linux-setup.c +22 -4
@@ -64,7 +64,12 @@ void setup_linux_bootloader_parameters(/* Find the maximum initial ramdisk address */ initrd_addr_max =DEFAULT_INITRD_ADDR_MAX ;if (real_mode->protocol_version >=0x0203 ) { - initrd_addr_max = real_mode->initrd_addr_max; +if (real_mode->protocol_version >=0x020c && + real_mode->xloadflags & (1 <<0 ) &&/* CAN_BE_LOADED_ABOVE_4G */ + real_mode->xloadflags & (1 <<15 ))/* USE_EXT_BOOT_PARAMS */ + initrd_addr_max =ULONG_MAX ; +else + initrd_addr_max = real_mode->initrd_addr_max;dbgprintf ("initrd_addr_max is 0x%lx\n" , initrd_addr_max); }
@@ -81,8 +86,16 @@ void setup_linux_bootloader_parameters( }/* Ramdisk address and size */ - real_mode->initrd_start = initrd_base; - real_mode->initrd_size = initrd_size; + real_mode->initrd_start = initrd_base &0xffffffffUL ; + real_mode->initrd_size = initrd_size &0xffffffffUL ; + +if (real_mode->protocol_version >=0x020c && + (initrd_base &0xffffffffUL ) != initrd_base) + real_mode->ext_ramdisk_image = initrd_base >>32 ; + +if (real_mode->protocol_version >=0x020c && + (initrd_size &0xffffffffUL ) != initrd_size) + real_mode->ext_ramdisk_size = initrd_size >>32 ;/* The location of the command line */ /* if (real_mode_base == 0x90000) { */
@@ -91,7 +104,12 @@ void setup_linux_bootloader_parameters(/* setup_move_size */ /* } */ if (real_mode->protocol_version >=0x0202 ) { - real_mode->cmd_line_ptr = real_mode_base + cmdline_offset; +unsigned long cmd_line_ptr = real_mode_base + cmdline_offset; + + real_mode->cmd_line_ptr = cmd_line_ptr &0xffffffffUL ; +if ((real_mode->protocol_version >=0x020c ) && + ((cmd_line_ptr &0xffffffffUL ) != cmd_line_ptr)) + real_mode->ext_cmd_line_ptr = cmd_line_ptr >>32 ; }/* Fill in the command line */
Signature · 7 lines
1.7.7 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v5 5/5 kexec, x86_64: Load bzImage64 above 4G
In reply to Yinghai Luneed to check xloadflags to see the bzImage is for 64bit relocatable.
-v2: add kexec-bzImage64.c according to Eric.
-v3: don't need to purgatory under 2g after Eric's change to purgatory code. -v4: use locate_hole find position first then add_buffer... suggested by Eric
add buffer for kernel image at last to make kexec-load faster. use xloadflags in setup_header to tell if is bzImage64. remove not cross GB boundary searching.
add --entry-32bit and --real-mode for skipping bzImage64.
--- kexec/arch/i386/include/arch/options.h | 4 +- kexec/arch/x86_64/Makefile | 1 + kexec/arch/x86_64/kexec-bzImage64.c | 335 ++++++++++++++++++++++++++++++++ kexec/arch/x86_64/kexec-x86_64.c | 1 + kexec/arch/x86_64/kexec-x86_64.h | 5 + 5 files changed, 345 insertions(+), 1 deletions(-) create mode 100644 kexec/arch/x86_64/kexec-bzImage64.c
kexec/arch/i386/include/arch/options.h +3 -1
@@ -29,6 +29,7 @@#define OPT_MOD (OPT_ARCH_MAX+7)#define OPT_VGA (OPT_ARCH_MAX+8)#define OPT_REAL_MODE (OPT_ARCH_MAX+9) +#define OPT_ENTRY_32BIT (OPT_ARCH_MAX+10)/* Options relevant to the architecture (excluding loader-specific ones): */ #define KEXEC_ARCH_OPTIONS \
@@ -68,7 +69,8 @@ { "args-linux", 0, NULL, OPT_ARGS_LINUX }, \ { "args-none", 0, NULL, OPT_ARGS_NONE }, \ { "module", 1, 0, OPT_MOD }, \ - { "real-mode", 0, NULL, OPT_REAL_MODE }, + { "real-mode", 0, NULL, OPT_REAL_MODE }, \ + { "entry-32bit", 0, NULL, OPT_ENTRY_32BIT }, #define KEXEC_ALL_OPT_STR KEXEC_ARCH_OPT_STR
kexec/arch/x86_64/Makefile +1 -0
@@ -13,6 +13,7 @@ x86_64_KEXEC_SRCS += kexec/arch/i386/crashdump-x86.c x86_64_KEXEC_SRCS_native = kexec/arch/x86_64/kexec-x86_64.c x86_64_KEXEC_SRCS_native += kexec/arch/x86_64/kexec-elf-x86_64.c x86_64_KEXEC_SRCS_native += kexec/arch/x86_64/kexec-elf-rel-x86_64.c +x86_64_KEXEC_SRCS_native += kexec/arch/x86_64/kexec-bzImage64.c x86_64_KEXEC_SRCS += $(x86_64_KEXEC_SRCS_native)
kexec/arch/x86_64/kexec-bzImage64.c new file +335 -0
335 lines not on this page · raw message
kexec/arch/x86_64/kexec-x86_64.c +1 -0
@@ -37,6 +37,7 @@ struct file_type file_type[] = { {"multiboot-x86" , multiboot_x86_probe, multiboot_x86_load, multiboot_x86_usage }, { "elf-x86", elf_x86_probe, elf_x86_load, elf_x86_usage }, + { "bzImage64", bzImage64_probe, bzImage64_load, bzImage64_usage }, { "bzImage", bzImage_probe, bzImage_load, bzImage_usage }, { "beoboot-x86", beoboot_probe, beoboot_load, beoboot_usage }, { "nbi-x86", nbi_probe, nbi_load, nbi_usage },
kexec/arch/x86_64/kexec-x86_64.h +5 -0
@@ -28,4 +28,9 @@ int elf_x86_64_load(int argc, char **argv, const char *buf, off_t len, struct kexec_info *info);void elf_x86_64_usage (void ); +int bzImage64_probe (const char *buf,off_t len); +int bzImage64_load (int argc,char **argv,const char *buf,off_t len, +struct kexec_info *info); +void bzImage64_usage (void ); +#endif /* KEXEC_X86_64_H */
Signature · 7 lines
1.7.7 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Hi Everyone,
there was some healthy discussion around the previous versions of this series, but so far none on this version. Have we reached consensus?
Quoted 23 lines · Yinghai Lu
Now we have limit kdump reserved under 896M, because kexec has the limitation. and also bzImage need to stay under 4g.
kernel parts changes could be found at:
git://git.kernel.org/pub/scm/linux/kernel/git/yinghai/linux-yinghai.gitfor-x86-boothere patches are for kexec tools to load bzImage and ramdisk above 4G acccording to new added boot header fields.
(quoted diff) +0
-3-v3: address review from Eric to use locate_hole at first. use xloadflags instead. -v4: remove the restriction about bzImage not crossing GB boundary. add real-mode fix for bzImage. add --entry-32bit and --real-mode for skip bzImage64. -v5: use USE_EXT_BOOT_PARAMS bit in xloadflags.Yinghai Lu (5):
kexec, x86: add boot header member for version 2.12 kexec, x86: cleanboot_paramsarea for entry-32bit path kexec, x86: Fix bzImage real-mode booting
kexec, x86: put ramdisk/cmd_line above 4G for 64bit bzImage kexec,x86_64: Load bzImage64 above 4G
Quoted 14 lines
include/x86/x86-linux.h | 32 +++- kexec/arch/i386/include/arch/options.h | 4 +- kexec/arch/i386/kexec-bzImage.c | 80 +++++++-- kexec/arch/i386/x86-linux-setup.c | 26 ++- kexec/arch/x86_64/Makefile | 1 + kexec/arch/x86_64/kexec-bzImage64.c | 335 ++++++++++++++++++++++++++++++++ kexec/arch/x86_64/kexec-x86_64.c | 1 + kexec/arch/x86_64/kexec-x86_64.h | 5 + 8 files changed, 462 insertions(+), 22 deletions(-) create mode 100644 kexec/arch/x86_64/kexec-bzImage64.cSignature · 1 line
1.7.7
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Quoted 6 lines · Simon Horman
Hi Everyone,
there was some healthy discussion around the previous versions of this series, but so far none on this version. Have we reached consensus?
No, we haven't converged on the boot protocol yet and just respinning the patch set doesn't change that.
-hpa
Signature · 8 lines
H. Peter Anvin, Intel Open Source Technology Center I work for Intel. I don't speak on their behalf. _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Quoted 11 lines · H. Peter Anvin
Quoted 7 lines · Simon Horman
Hi Everyone,
there was some healthy discussion around the previous versions of this series, but so far none on this version. Have we reached consensus?
No, we haven't converged on the boot protocol yet and just respinning the patch set doesn't change that.
please check if you are ok with bit 15 in xloadflags.
+Field name: xloadflags +Type: modify (obligatory) +Offset/size: 0x236/2 +Protocol: 2.12+
+
+ This field is a bitmask.
+
+ Bit 0 (read): CAN_BE_LOADED_ABOVE_4G
+ - If 1, kernel/boot_params/cmdline/ramdisk can be above 4g,
+
+ Bit 15 (write): USE_EXT_BOOT_PARAMS
+ - If 1, set by bootloader, and kernel could check new fields
+ in boot_params that are added from 2.12 safely.
Thanks
Yinghai
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec