x86/boot: fix relying on link order
cover x86/boot: fix relying on link order
The series contains stuff I caught last week while working on some x86 code. It can be considered a material for fixes or for next, I'm okay with either, although leaning more towards fixes :P
Notes on patches:
* 0001: I didn't put any "Fixes:" tag since it's not linear. The lines changed with this patch came from the initial x86 KASLR series, but that unconditional jump to the kernel beginning already was there. It goes at least from the set that brought relocatable kernel support to x86, but this is quite prehistoric already and might not look really relatable. So up to you whether it needs any.
* 0002: doesn't fix anything, except that having any files listed in that whitelist already is an architecture bug :D And it wouldn't be convenient to decouple it from #0001, but up to you as well.
Alexander Lobakin (2):
x86/boot: robustify calling startup_{32,64}() from the decompressor
code
scripts/head-object-list: remove x86 from the list
arch/x86/boot/compressed/head_32.S | 2 +- arch/x86/boot/compressed/head_64.S | 2 +- arch/x86/boot/compressed/misc.c | 16 +++++++++++----- scripts/head-object-list.txt | 6 ------ 4 files changed, 13 insertions(+), 13 deletions(-)
Signature · 1 line
2.38.1
PATCH 1/2 x86/boot: robustify calling startup_{32,64}() from the decompressor code
After commit ce697ccee1a8 ("kbuild: remove head-y syntax"), I started digging whether x86 is ready from removing this old cruft. Removing its objects from the list makes the kernel unbootable. This applies only to bzImage, vmlinux still works correctly. The reason is that with no strict object order determined by the linker arguments, not the linker script, startup_64 can be placed not right at the beginning of the kernel.
Here's vmlinux.map's beginning before removing:
ffffffff81000000 vmlinux.o:(.head.text)
ffffffff81000000 startup_64 ffffffff81000070 secondary_startup_64 ffffffff81000075 secondary_startup_64_no_verify ffffffff81000160 verify_cpu
and after:
ffffffff81000000 vmlinux.o:(.head.text)
ffffffff81000000 pvh_start_xen ffffffff81000080 startup_64 ffffffff810000f0 secondary_startup_64 ffffffff810000f5 secondary_startup_64_no_verify
Not a problem itself, but the self-extractor code has the address of that function hardcoded the beginning, not looking onto the ELF header, which always contains the address of startup_{32,64}().
So, instead of doing an "act of blind faith", just take the address from the ELF header and extract a relative offset to the entry point. The decompressor function already returns a pointer to the beginning of the kernel to the Asm code, which then jumps to it, so add that offset to the return value.
This doesn't change anything for now, but allows to resign from the "head object list" for x86 and makes sure valid Kbuild or any other improvements won't break anything here in general.
--- arch/x86/boot/compressed/head_32.S | 2 +- arch/x86/boot/compressed/head_64.S | 2 +- arch/x86/boot/compressed/misc.c | 16 +++++++++++----- 3 files changed, 13 insertions(+), 7 deletions(-)
arch/x86/boot/compressed/head_32.S +1 -1
@@ -187,7 +187,7 @@ SYM_FUNC_START_LOCAL_NOALIGN(.Lrelocated) leal boot_heap@GOTOFF(%ebx), %eax pushl %eax /* heap area */ pushl %esi /* real mode pointer */ - call extract_kernel /* returns kernel location in %eax */ + call extract_kernel /* returns kernel entry point in %eax */ addl $24, %esp /*
arch/x86/boot/compressed/head_64.S +1 -1
@@ -593,7 +593,7 @@ SYM_FUNC_START_LOCAL_NOALIGN(.Lrelocated) movl input_len(%rip), %ecx /* input_len */ movq %rbp, %r8 /* output target address */ movl output_len(%rip), %r9d /* decompressed length, end of relocs */ - call extract_kernel /* returns kernel location in %rax */ + call extract_kernel /* returns kernel entry point in %rax */ popq %rsi /*
arch/x86/boot/compressed/misc.c +11 -5
@@ -277,7 +277,7 @@ static inline void handle_relocations(void *output, unsigned long output_len, { }#endif -static void parse_elf (void *output) +static size_t parse_elf (void *output) {#ifdef CONFIG_X86_64 Elf64_Ehdr ehdr;
@@ -287,6 +287,7 @@ static void parse_elf(void *output)Elf32_Phdr *phdrs, *phdr;#endif void *dest; +size_t off;int i;memcpy (&ehdr, output,sizeof (ehdr));
@@ -295,16 +296,19 @@ static void parse_elf(void *output) ehdr.e_ident[EI_MAG2] != ELFMAG2 || ehdr.e_ident[EI_MAG3] != ELFMAG3) {error ("Kernel is not a valid ELF file" ); -return ; +return 0 ; }debug_putstr ("Parsing ELF... " ); phdrs =malloc (sizeof (*phdrs) * ehdr.e_phnum); -if (!phdrs) +if (!phdrs) {error ("Failed to allocate space for phdrs" ); +return 0 ; + }memcpy (phdrs, output + ehdr.e_phoff,sizeof (*phdrs) * ehdr.e_phnum); + off = ehdr.e_entry - phdrs->p_paddr;for (i =0 ; i < ehdr.e_phnum; i++) { phdr = &phdrs[i];
@@ -328,6 +332,7 @@ static void parse_elf(void *output) }free (phdrs); +return off; }/*
@@ -356,6 +361,7 @@ asmlinkage __visible void *extract_kernel(void *rmode, memptr heap,const unsigned long kernel_total_size = VO__end - VO__text;unsigned long virt_addr =LOAD_PHYSICAL_ADDR ;unsigned long needed_size; +size_t off;/* Retain x86 boot parameters pointer passed from startup_32/64. */ boot_params = rmode;
@@ -456,14 +462,14 @@ asmlinkage __visible void *extract_kernel(void *rmode, memptr heap,debug_putstr ("\nDecompressing Linux... " );__decompress (input_data, input_len,NULL ,NULL , output, output_len,NULL , error); -parse_elf (output); + off =parse_elf (output);handle_relocations (output, output_len, virt_addr);debug_putstr ("done.\nBooting the kernel.\n" );/* Disable exception handling before booting the kernel */ cleanup_exception_handling (); -return output; +return output + off; }void fortify_panic (const char *name)
Signature · 1 line
2.38.1
H. Peter Anvin
reviewerRe: [PATCH 1/2] x86/boot: robustify calling startup_{32,64}() from the decompressor code
Quoted 135 lines · Alexander Lobakin
After commit
ce697ccee1a8("kbuild: remove head-y syntax"), I started digging whether x86 is ready from removing this old cruft. Removing its objects from the list makes the kernel unbootable. This applies only to bzImage, vmlinux still works correctly. The reason is that with no strict object order determined by the linker arguments, not the linker script,startup_64can be placed not right at the beginning of the kernel.
Here's vmlinux.map's beginning before removing:
ffffffff81000000vmlinux.o:(.head.text)ffffffff81000000 startup_64 ffffffff81000070 secondary_startup_64 ffffffff81000075 secondary_startup_64_no_verify ffffffff81000160 verify_cpuand after:
ffffffff81000000vmlinux.o:(.head.text)ffffffff81000000 pvh_start_xen ffffffff81000080 startup_64 ffffffff810000f0 secondary_startup_64 ffffffff810000f5 secondary_startup_64_no_verifyNot a problem itself, but the self-extractor code has the address of that function hardcoded the beginning, not looking onto the ELF header, which always contains the address of startup_{32,64}().
So, instead of doing an "act of blind faith", just take the address from the ELF header and extract a relative offset to the entry point. The decompressor function already returns a pointer to the beginning of the kernel to the Asm code, which then jumps to it, so add that offset to the return value.
This doesn't change anything for now, but allows to resign from the "head object list" for x86 and makes sure validKbuildor any other improvements won't break anything here in general.Signed-off-by Alexander Lobakin <[email protected]>---
arch/x86/boot/compressed/head_32.S | 2 +- arch/x86/boot/compressed/head_64.S | 2 +- arch/x86/boot/compressed/misc.c | 16 +++++++++++----- 3 files changed, 13 insertions(+), 7 deletions(-)arch/x86/boot/compressed/head_32.S +1
-1@@ -187,7 +187,7 @@ SYM_FUNC_START_LOCAL_NOALIGN(.Lrelocated) leal boot_heap@GOTOFF(%ebx), %eax pushl %eax /* heap area */ pushl %esi /* real mode pointer */ - call extract_kernel /* returns kernel location in %eax */ + call extract_kernel /* returns kernel entry point in %eax */ addl $24, %esp /*arch/x86/boot/compressed/head_64.S +1
-1@@ -593,7 +593,7 @@ SYM_FUNC_START_LOCAL_NOALIGN(.Lrelocated) movl input_len(%rip), %ecx /* input_len */ movq %rbp, %r8 /* output target address */ movl output_len(%rip), %r9d /* decompressed length, end of relocs */ - call extract_kernel /* returns kernel location in %rax */ + call extract_kernel /* returns kernel entry point in %rax */ popq %rsi /*arch/x86/boot/compressed/misc.c +10
-4@@ -277,7 +277,7 @@ static inline void handle_relocations(void *output, unsigned long output_len, { }#endif -static void parse_elf (void *output) +static size_t parse_elf (void *output) {#ifdef CONFIG_X86_64 Elf64_Ehdr ehdr;@@ -287,6 +287,7 @@ static void parse_elf(void *output)Elf32_Phdr *phdrs, *phdr;#endif void *dest; +size_t off;int i;memcpy (&ehdr, output,sizeof (ehdr));@@ -295,16 +296,19 @@ static void parse_elf(void *output) ehdr.e_ident[EI_MAG2] != ELFMAG2 || ehdr.e_ident[EI_MAG3] != ELFMAG3) {error ("Kernel is not a valid ELF file" ); -return ; +return 0 ; }debug_putstr ("Parsing ELF... " ); phdrs =malloc (sizeof (*phdrs) * ehdr.e_phnum); -if (!phdrs) +if (!phdrs) {error ("Failed to allocate space for phdrs" ); +return 0 ; + }memcpy (phdrs, output + ehdr.e_phoff,sizeof (*phdrs) * ehdr.e_phnum); + off = ehdr.e_entry - phdrs->p_paddr;for (i =0 ; i < ehdr.e_phnum; i++) { phdr = &phdrs[i];@@ -328,6 +332,7 @@ static void parse_elf(void *output) }free (phdrs); +return off; }/* @@ -356,6 +361,7 @@ asmlinkage __visible void *extract_kernel(void *rmode, memptr heap,const unsigned long kernel_total_size = VO__end - VO__text;unsigned long virt_addr =LOAD_PHYSICAL_ADDR ;unsigned long needed_size; +size_t off;/* Retain x86 boot parameters pointer passed from startup_32/64. */ boot_params = rmode;@@ -456,14 +462,14 @@ asmlinkage __visible void *extract_kernel(void *rmode, memptr heap,debug_putstr ("\nDecompressing Linux... " );__decompress (input_data, input_len,NULL ,NULL , output, output_len,NULL , error); -parse_elf (output); + off =parse_elf (output);handle_relocations (output, output_len, virt_addr);
arch/x86/boot/compressed/misc.c asmlinkage __visible void *extract_kernel(void *rmode, memptr heap, · line 475
Quoted 10 lines
arch/x86/boot/compressed/misc.c +1
-1debug_putstr ("done.\nBooting the kernel.\n" );/* Disable exception handling before booting the kernel */ cleanup_exception_handling (); -return output; +return output + off; }void fortify_panic (const char *name)
Put the entry points in dedicated sections and declare them explicitly in the linker script (the standard name for this section would be .init, but we presumably want .init.32 and .init.64 to keep 32- and 64-bit code in separate sections.)
Re: [PATCH 1/2] x86/boot: robustify calling startup_{32,64}() from the decompressor code
From: "H. Peter Anvin" <[email protected]>
Date: Mon, 31 Oct 2022 13:31:05 -0700
Quoted 9 lines · Alexander Lobakin
Quoted 8 lines
After commit
ce697ccee1a8("kbuild: remove head-y syntax"), I started digging whether x86 is ready from removing this old cruft. Removing its objects from the list makes the kernel unbootable. This applies only to bzImage, vmlinux still works correctly. The reason is that with no strict object order determined by the linker arguments, not the linker script,startup_64can be placed not right at the beginning of the kernel.
Here's vmlinux.map's beginning before removing:
[...]
Quoted 7 lines
Quoted 5 lines
arch/x86/boot/compressed/misc.c +1
-1-return output; +return output + off;}
void
fortify_panic(const char *name)Put the entry points in dedicated sections and declare them explicitly in the linker script (the standard name for this section would be .init, but we presumably want .init.32 and .init.64 to keep 32- and 64-bit code in separate sections.)
Uhm, I'm not sure I follow. We're trying to eliminate as much hardcode as possible, and now you propose to introduce new :P Some architectures don't have a separate .head.text at all: startup functions are placed in .init.text, preboot code finds the entry point and it just works. It's not possible currently on x86_64, but we'll be there one day.
Thanks,
Olek
Re: [PATCH 1/2] x86/boot: robustify calling startup_{32,64}() from the decompressor code
In reply to Alexander LobakinQuoted 27 lines · Alexander Lobakin
After commit
ce697ccee1a8("kbuild: remove head-y syntax"), I started digging whether x86 is ready from removing this old cruft. Removing its objects from the list makes the kernel unbootable. This applies only to bzImage, vmlinux still works correctly. The reason is that with no strict object order determined by the linker arguments, not the linker script,startup_64can be placed not right at the beginning of the kernel.
Here's vmlinux.map's beginning before removing:
ffffffff81000000vmlinux.o:(.head.text)ffffffff81000000 startup_64 ffffffff81000070 secondary_startup_64 ffffffff81000075 secondary_startup_64_no_verify ffffffff81000160 verify_cpuand after:
ffffffff81000000vmlinux.o:(.head.text)ffffffff81000000 pvh_start_xen ffffffff81000080 startup_64 ffffffff810000f0 secondary_startup_64 ffffffff810000f5 secondary_startup_64_no_verifyNot a problem itself, but the self-extractor code has the address of that function hardcoded the beginning, not looking onto the ELF header, which always contains the address of startup_{32,64}().
Quoted 8 lines
So, instead of doing an "act of blind faith", just take the address from the ELF header and extract a relative offset to the entry point. The decompressor function already returns a pointer to the beginning of the kernel to the Asm code, which then jumps to it, so add that offset to the return value.
This doesn't change anything for now, but allows to resign from the "head object list" for x86 and makes sure validKbuildor any other improvements won't break anything here in general.
Oh yeah! I wouldn't think that implementing this would be _that_ easy.
The next natural step would be to eliminate the whole head section. But that would need a bit more work as not all jumps are rip-relative, apparently...
Few comments below, so no Reviewed-by yet.
Quoted 1 line
Signed-off-by Alexander Lobakin <[email protected]>
...
Quoted 20 lines
arch/x86/boot/compressed/misc.c +2
-1@@ -277,7 +277,7 @@ static inline void handle_relocations(void *output, unsigned long output_len, { }#endif -static void parse_elf (void *output) +static size_t parse_elf (void *output) {#ifdef CONFIG_X86_64 Elf64_Ehdr ehdr;@@ -287,6 +287,7 @@ static void parse_elf(void *output)Elf32_Phdr *phdrs, *phdr;#endif void *dest; +size_t off;int i;memcpy (&ehdr, output,sizeof (ehdr));
arch/x86/boot/compressed/misc.c static void parse_elf(void *output) · line 299
Quoted 5 lines
arch/x86/boot/compressed/misc.c +1
-1@@ -295,16 +296,19 @@ static void parse_elf(void *output) ehdr.e_ident[EI_MAG2] != ELFMAG2 || ehdr.e_ident[EI_MAG3] != ELFMAG3) {error ("Kernel is not a valid ELF file" ); -return ; +return 0 ;
error() is noreturn, so you can remove these returns. They don't make sense anyway. Likely in a separate patch.
Quoted 17 lines
arch/x86/boot/compressed/misc.c +4
-1}debug_putstr ("Parsing ELF... " ); phdrs =malloc (sizeof (*phdrs) * ehdr.e_phnum); -if (!phdrs) +if (!phdrs) {error ("Failed to allocate space for phdrs" ); +return 0 ; + }memcpy (phdrs, output + ehdr.e_phoff,sizeof (*phdrs) * ehdr.e_phnum); + off = ehdr.e_entry - phdrs->p_paddr;for (i =0 ; i < ehdr.e_phnum; i++) { phdr = &phdrs[i];
arch/x86/boot/compressed/misc.c static void parse_elf(void *output) · line 335
Quoted 4 lines
arch/x86/boot/compressed/misc.c +1
-0@@ -328,6 +332,7 @@ static void parse_elf(void *output) }free (phdrs); +return off;
You should add a \n before the return.
Quoted 12 lines
arch/x86/boot/compressed/misc.c +1
-0}/* @@ -356,6 +361,7 @@ asmlinkage __visible void *extract_kernel(void *rmode, memptr heap,const unsigned long kernel_total_size = VO__end - VO__text;unsigned long virt_addr =LOAD_PHYSICAL_ADDR ;unsigned long needed_size; +size_t off;/* Retain x86 boot parameters pointer passed from startup_32/64. */ boot_params = rmode;
arch/x86/boot/compressed/misc.c asmlinkage __visible void *extract_kernel(void *rmode, memptr heap, · line 465
Quoted 5 lines
arch/x86/boot/compressed/misc.c +1
-1@@ -456,14 +462,14 @@ asmlinkage __visible void *extract_kernel(void *rmode, memptr heap,debug_putstr ("\nDecompressing Linux... " );__decompress (input_data, input_len,NULL ,NULL , output, output_len,NULL , error); -parse_elf (output); + off =parse_elf (output);
Perhaps add:
debug_putaddr(off);
here?
Quoted 11 lines
arch/x86/boot/compressed/misc.c +1
-1handle_relocations (output, output_len, virt_addr);debug_putstr ("done.\nBooting the kernel.\n" );/* Disable exception handling before booting the kernel */ cleanup_exception_handling (); -return output; +return output + off; }
Quoted 1 line
void
fortify_panic(const char *name)
thanks,
Signature · 2 lines
js suse labs
Re: [PATCH 1/2] x86/boot: robustify calling startup_{32,64}() from the decompressor code
Quoted 1 line · Jiri Slaby
Tested-by Jiri Slaby <[email protected]>
FWIW that means: tested with gcc-lto which deliberately stores startup_64() randomly, not to the beginning of vmlinux.
Signature · 2 lines
js suse labs
Re: [PATCH 1/2] x86/boot: robustify calling startup_{32,64}() from the decompressor code
From: Jiri Slaby <[email protected]>
Date: Tue, 1 Nov 2022 07:50:04 +0100
Quoted 4 lines · Jiri Slaby
Quoted 1 line
Tested-by Jiri Slaby <[email protected]>FWIW that means: tested with gcc-lto which deliberately stores
startup_64()randomly, not to the beginning of vmlinux.
Yeah, you never know. After rebasing on top of 6.1-rc1 my FG-KASLR series stopped booting, I found out that the combo of that head-object-list.txt + -ffunction-sections (but for Asm) also made startup_64() appear in random parts of .head.text.
Re head removal -- that would be awesome, but currently from what I've tried during the development, there's a lot of stuff to be redone, the compressor code sometimes just doesn't want to work if you play with .head.text too much. Will see.
Quoted 4 lines
Signature · 3 lines
-- js suse labs
Thanks,
Olek
PATCH 2/2 scripts/head-object-list: remove x86 from the list
In reply to Alexander LobakinNow that x86 boot code is not hardcoded to the particular linking order, remove x86 files from the list and let them be placed inside the vmlinux according only to the linker script and linker preferences.
--- scripts/head-object-list.txt | 6 ------ 1 file changed, 6 deletions(-)
scripts/head-object-list.txt +0 -6
@@ -44,10 +44,4 @@ arch/s390/kernel/head64.o arch/sh/kernel/head_32.o arch/sparc/kernel/head_32.o arch/sparc/kernel/head_64.o -arch/x86/kernel/head_32.o -arch/x86/kernel/head_64.o -arch/x86/kernel/head32.o -arch/x86/kernel/head64.o -arch/x86/kernel/ebda.o -arch/x86/kernel/platform-quirks.o arch/xtensa/kernel/head.o
Signature · 1 line
2.38.1
Re: [PATCH 2/2] scripts/head-object-list: remove x86 from the list
Quoted 6 lines · Alexander Lobakin
Now that x86 boot code is not hardcoded to the particular linking order, remove x86 files from the list and let them be placed inside the vmlinux according only to the linker script and linker preferences.
Signed-off-by Alexander Lobakin <[email protected]>
Quoted 19 lines
--- scripts/head-object-list.txt | 6 ------ 1 file changed, 6 deletions(-)scripts/head-object-list.txt +0
-6@@ -44,10 +44,4 @@ arch/s390/kernel/head64.o arch/sh/kernel/head_32.o arch/sparc/kernel/head_32.o arch/sparc/kernel/head_64.o -arch/x86/kernel/head_32.o -arch/x86/kernel/head_64.o -arch/x86/kernel/head32.o -arch/x86/kernel/head64.o -arch/x86/kernel/ebda.o -arch/x86/kernel/platform-quirks.o arch/xtensa/kernel/head.o
Signature · 2 lines
js suse labs
Borislav Petkov
maintainer+ jslaby.
Quoted 31 lines · Alexander Lobakin
The series contains stuff I caught last week while working on some x86 code. It can be considered a material for fixes or for next, I'm okay with either, although leaning more towards fixes :P
Notes on patches:
* 0001: I didn't put any "Fixes:" tag since it's not linear. The lines changed with this patch came from the initial x86 KASLR series, but that unconditional jump to the kernel beginning already was there. It goes at least from the set that brought relocatable kernel support to x86, but this is quite prehistoric already and might not look really relatable. So up to you whether it needs any.
* 0002: doesn't fix anything, except that having any files listed in that whitelist already is an architecture bug :D And it wouldn't be convenient to decouple it from #0001, but up to you as well.Alexander Lobakin (2):
x86/boot: robustify calling startup_{32,64}() from the decompressor
code
scripts/head-object-list: remove x86 from the listarch/x86/boot/compressed/head_32.S | 2 +- arch/x86/boot/compressed/head_64.S | 2 +- arch/x86/boot/compressed/misc.c | 16 +++++++++++----- scripts/head-object-list.txt | 6 ------ 4 files changed, 13 insertions(+), 13 deletions(-)Signature · 1 line
2.38.1
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette