| Message ID | 1600959521-24158-8-git-send-email-ross.philipson@oracle.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | x86: Trenchboot secure dynamic launch Linux kernel support | expand |
On Thu, Sep 24, 2020 at 10:58:35AM -0400, Ross Philipson wrote: > The Secure Launch (SL) stub provides the entry point for Intel TXT (and > later AMD SKINIT) to vector to during the late launch. The symbol > sl_stub_entry is that entry point and its offset into the kernel is > conveyed to the launching code using the MLE (Measured Launch > Environment) header in the structure named mle_header. The offset of the > MLE header is set in the kernel_info. The routine sl_stub contains the > very early late launch setup code responsible for setting up the basic > environment to allow the normal kernel startup_32 code to proceed. It is > also responsible for properly waking and handling the APs on Intel > platforms. The routine sl_main which runs after entering 64b mode is > responsible for measuring configuration and module information before > it is used like the boot params, the kernel command line, the TXT heap, > an external initramfs, etc. > > Signed-off-by: Ross Philipson <ross.philipson@oracle.com> Which version of the kernel is this based on? > diff --git a/arch/x86/boot/compressed/head_64.S b/arch/x86/boot/compressed/head_64.S > index 97d37f0..42043bf 100644 > --- a/arch/x86/boot/compressed/head_64.S > +++ b/arch/x86/boot/compressed/head_64.S > @@ -279,6 +279,21 @@ SYM_INNER_LABEL(efi32_pe_stub_entry, SYM_L_LOCAL) > SYM_FUNC_END(efi32_stub_entry) > #endif > > +#ifdef CONFIG_SECURE_LAUNCH > +SYM_FUNC_START(sl_stub_entry) > + /* > + * On entry, %ebx has the entry abs offset to sl_stub_entry. To > + * find the beginning of where we are loaded, sub off from the > + * beginning. > + */ This requirement should be added to the documentation. Is it necessary or can this stub just figure out the address the same way as the other 32-bit entry points, using the scratch space in bootparams as a little stack? > + leal (startup_32 - sl_stub_entry)(%ebx), %ebx > + > + /* More room to work in sl_stub in the text section */ > + jmp sl_stub > + > +SYM_FUNC_END(sl_stub_entry) > +#endif > + > .code64 > .org 0x200 > SYM_CODE_START(startup_64) > @@ -537,6 +552,25 @@ SYM_FUNC_START_LOCAL_NOALIGN(.Lrelocated) > shrq $3, %rcx > rep stosq > > +#ifdef CONFIG_SECURE_LAUNCH > + /* > + * Have to do the final early sl stub work in 64b area. > + * > + * *********** NOTE *********** > + * > + * Several boot params get used before we get a chance to measure > + * them in this call. This is a known issue and we currently don't > + * have a solution. The scratch field doesn't matter and loadflags > + * have KEEP_SEGMENTS set by the stub code. There is no obvious way > + * to do anything about the use of kernel_alignment or init_size > + * though these seem low risk. > + */ There are various fields in bootparams that depend on where the kernel/initrd and cmdline are loaded in memory. If the entire bootparams page is getting measured, does that mean they all have to be at fixed addresses on every boot? Also KEEP_SEGMENTS support is gone from the kernel since v5.7, since it was unused. startup_32 now always loads a GDT and then the segment registers. I think this should be ok for you as the only thing the flag used to do in the 64-bit kernel was to stop startup_32 from blindly loading __BOOT_DS into the segment registers before it had setup its own GDT. For the 32-bit assembler code that's being added, tip/master now has changes that prevent the compressed kernel from having any runtime relocations. You'll need to revise some of the code and the data structures initial values to avoid creating relocations. Thanks.
On 9/24/20 1:38 PM, Arvind Sankar wrote: > On Thu, Sep 24, 2020 at 10:58:35AM -0400, Ross Philipson wrote: >> The Secure Launch (SL) stub provides the entry point for Intel TXT (and >> later AMD SKINIT) to vector to during the late launch. The symbol >> sl_stub_entry is that entry point and its offset into the kernel is >> conveyed to the launching code using the MLE (Measured Launch >> Environment) header in the structure named mle_header. The offset of the >> MLE header is set in the kernel_info. The routine sl_stub contains the >> very early late launch setup code responsible for setting up the basic >> environment to allow the normal kernel startup_32 code to proceed. It is >> also responsible for properly waking and handling the APs on Intel >> platforms. The routine sl_main which runs after entering 64b mode is >> responsible for measuring configuration and module information before >> it is used like the boot params, the kernel command line, the TXT heap, >> an external initramfs, etc. >> >> Signed-off-by: Ross Philipson <ross.philipson@oracle.com> > > Which version of the kernel is this based on? git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master branch > >> diff --git a/arch/x86/boot/compressed/head_64.S b/arch/x86/boot/compressed/head_64.S >> index 97d37f0..42043bf 100644 >> --- a/arch/x86/boot/compressed/head_64.S >> +++ b/arch/x86/boot/compressed/head_64.S >> @@ -279,6 +279,21 @@ SYM_INNER_LABEL(efi32_pe_stub_entry, SYM_L_LOCAL) >> SYM_FUNC_END(efi32_stub_entry) >> #endif >> >> +#ifdef CONFIG_SECURE_LAUNCH >> +SYM_FUNC_START(sl_stub_entry) >> + /* >> + * On entry, %ebx has the entry abs offset to sl_stub_entry. To >> + * find the beginning of where we are loaded, sub off from the >> + * beginning. >> + */ > > This requirement should be added to the documentation. Is it necessary > or can this stub just figure out the address the same way as the other > 32-bit entry points, using the scratch space in bootparams as a little > stack? It is based on the state of the BSP when TXT vectors to the measured launch environment. It is documented in the TXT spec and the SDMs. > >> + leal (startup_32 - sl_stub_entry)(%ebx), %ebx >> + >> + /* More room to work in sl_stub in the text section */ >> + jmp sl_stub >> + >> +SYM_FUNC_END(sl_stub_entry) >> +#endif >> + >> .code64 >> .org 0x200 >> SYM_CODE_START(startup_64) >> @@ -537,6 +552,25 @@ SYM_FUNC_START_LOCAL_NOALIGN(.Lrelocated) >> shrq $3, %rcx >> rep stosq >> >> +#ifdef CONFIG_SECURE_LAUNCH >> + /* >> + * Have to do the final early sl stub work in 64b area. >> + * >> + * *********** NOTE *********** >> + * >> + * Several boot params get used before we get a chance to measure >> + * them in this call. This is a known issue and we currently don't >> + * have a solution. The scratch field doesn't matter and loadflags >> + * have KEEP_SEGMENTS set by the stub code. There is no obvious way >> + * to do anything about the use of kernel_alignment or init_size >> + * though these seem low risk. >> + */ > > There are various fields in bootparams that depend on where the > kernel/initrd and cmdline are loaded in memory. If the entire bootparams > page is getting measured, does that mean they all have to be at fixed > addresses on every boot? Yes that is a very good point. In other places when measuring we make sure to skip things like addresses and sizes of things outside of the structure being measured. This needs to be done with boot params too. > > Also KEEP_SEGMENTS support is gone from the kernel since v5.7, since it > was unused. startup_32 now always loads a GDT and then the segment > registers. I think this should be ok for you as the only thing the flag > used to do in the 64-bit kernel was to stop startup_32 from blindly > loading __BOOT_DS into the segment registers before it had setup its own > GDT. Yea this was there to prevent that blind loading of __BOOT_DS. I see it is gone so I will remove the comment and the place where the flag is set. > > For the 32-bit assembler code that's being added, tip/master now has > changes that prevent the compressed kernel from having any runtime > relocations. You'll need to revise some of the code and the data > structures initial values to avoid creating relocations. Could you elaborate on this some more? I am not sure I see places in the secure launch asm that would be creating relocations like this. Thank you, Ross > > Thanks. >
On Fri, Sep 25, 2020 at 10:56:43AM -0400, Ross Philipson wrote: > On 9/24/20 1:38 PM, Arvind Sankar wrote: > > On Thu, Sep 24, 2020 at 10:58:35AM -0400, Ross Philipson wrote: > > > >> diff --git a/arch/x86/boot/compressed/head_64.S b/arch/x86/boot/compressed/head_64.S > >> index 97d37f0..42043bf 100644 > >> --- a/arch/x86/boot/compressed/head_64.S > >> +++ b/arch/x86/boot/compressed/head_64.S > >> @@ -279,6 +279,21 @@ SYM_INNER_LABEL(efi32_pe_stub_entry, SYM_L_LOCAL) > >> SYM_FUNC_END(efi32_stub_entry) > >> #endif > >> > >> +#ifdef CONFIG_SECURE_LAUNCH > >> +SYM_FUNC_START(sl_stub_entry) > >> + /* > >> + * On entry, %ebx has the entry abs offset to sl_stub_entry. To > >> + * find the beginning of where we are loaded, sub off from the > >> + * beginning. > >> + */ > > > > This requirement should be added to the documentation. Is it necessary > > or can this stub just figure out the address the same way as the other > > 32-bit entry points, using the scratch space in bootparams as a little > > stack? > > It is based on the state of the BSP when TXT vectors to the measured > launch environment. It is documented in the TXT spec and the SDMs. > I think it would be useful to add to the x86 boot documentation how exactly this new entry point is called, even if it's just adding a link to some section of those specs. The doc should also say that an mle_header_offset of 0 means the kernel isn't secure launch enabled. > > > > For the 32-bit assembler code that's being added, tip/master now has > > changes that prevent the compressed kernel from having any runtime > > relocations. You'll need to revise some of the code and the data > > structures initial values to avoid creating relocations. > > Could you elaborate on this some more? I am not sure I see places in the > secure launch asm that would be creating relocations like this. > > Thank you, > Ross > You should see them if you do readelf -r arch/x86/boot/compressed/vmlinux In terms of the code, things like: addl %ebx, (sl_gdt_desc + 2)(%ebx) will create a relocation, because the linker interprets this as wanting the runtime address of sl_gdt_desc, rather than just the offset from startup_32. https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/tree/arch/x86/boot/compressed/head_64.S#n48 has a comment with some explanation and a macro that the 32-bit code in startup_32 uses to avoid creating relocations. Since the SL code is in a different assembler file (and a different section), you can't directly use the same macro. I would suggest getting rid of sl_stub_entry and entering directly at sl_stub, and then the code in sl_stub.S can use sl_stub for the base address, defining the rva() macro there as #define rva(X) ((X) - sl_stub) You will also need to avoid initializing data with symbol addresses. .long mle_header .long sl_stub_entry .long sl_gdt will create relocations. The third one is easy, just replace it with sl_gdt - sl_gdt_desc and initialize it at runtime with leal rva(sl_gdt_desc)(%ebx), %eax addl %eax, 2(%eax) lgdt (%eax) The other two are more messy, unfortunately there is no easy way to tell the linker what we want here. The other entry point addresses (for the EFI stub) are populated in a post-processing step after the compressed kernel has been linked, we could teach it to also update kernel_info. Without that, for kernel_info, you could change it to store the offset of the MLE header from kernel_info, instead of from the start of the image. For the MLE header, it could be moved to .head.text in head_64.S, and initialized with .long rva(sl_stub) This will also let it be placed at a fixed offset from startup_32, so that kernel_info can just be populated with a constant.
On 9/25/20 3:18 PM, Arvind Sankar wrote: > On Fri, Sep 25, 2020 at 10:56:43AM -0400, Ross Philipson wrote: >> On 9/24/20 1:38 PM, Arvind Sankar wrote: >>> On Thu, Sep 24, 2020 at 10:58:35AM -0400, Ross Philipson wrote: >>> >>>> diff --git a/arch/x86/boot/compressed/head_64.S b/arch/x86/boot/compressed/head_64.S >>>> index 97d37f0..42043bf 100644 >>>> --- a/arch/x86/boot/compressed/head_64.S >>>> +++ b/arch/x86/boot/compressed/head_64.S >>>> @@ -279,6 +279,21 @@ SYM_INNER_LABEL(efi32_pe_stub_entry, SYM_L_LOCAL) >>>> SYM_FUNC_END(efi32_stub_entry) >>>> #endif >>>> >>>> +#ifdef CONFIG_SECURE_LAUNCH >>>> +SYM_FUNC_START(sl_stub_entry) >>>> + /* >>>> + * On entry, %ebx has the entry abs offset to sl_stub_entry. To >>>> + * find the beginning of where we are loaded, sub off from the >>>> + * beginning. >>>> + */ >>> >>> This requirement should be added to the documentation. Is it necessary >>> or can this stub just figure out the address the same way as the other >>> 32-bit entry points, using the scratch space in bootparams as a little >>> stack? >> >> It is based on the state of the BSP when TXT vectors to the measured >> launch environment. It is documented in the TXT spec and the SDMs. >> > > I think it would be useful to add to the x86 boot documentation how > exactly this new entry point is called, even if it's just adding a link > to some section of those specs. The doc should also say that an > mle_header_offset of 0 means the kernel isn't secure launch enabled. Ok will do. > >>> >>> For the 32-bit assembler code that's being added, tip/master now has >>> changes that prevent the compressed kernel from having any runtime >>> relocations. You'll need to revise some of the code and the data >>> structures initial values to avoid creating relocations. >> >> Could you elaborate on this some more? I am not sure I see places in the >> secure launch asm that would be creating relocations like this. >> >> Thank you, >> Ross >> > > You should see them if you do > readelf -r arch/x86/boot/compressed/vmlinux > > In terms of the code, things like: > > addl %ebx, (sl_gdt_desc + 2)(%ebx) > > will create a relocation, because the linker interprets this as wanting > the runtime address of sl_gdt_desc, rather than just the offset from > startup_32. > > https://urldefense.com/v3/__https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/tree/arch/x86/boot/compressed/head_64.S*n48__;Iw!!GqivPVa7Brio!JpZWv1cCPZdjD2jbCCGT7P9UIVl_lhX7YjckAnUcvi927jwZI7X3nX0MpIAZOyktJds$ > > has a comment with some explanation and a macro that the 32-bit code in > startup_32 uses to avoid creating relocations. > > Since the SL code is in a different assembler file (and a different > section), you can't directly use the same macro. I would suggest getting > rid of sl_stub_entry and entering directly at sl_stub, and then the code > in sl_stub.S can use sl_stub for the base address, defining the rva() > macro there as > > #define rva(X) ((X) - sl_stub) > > You will also need to avoid initializing data with symbol addresses. > > .long mle_header > .long sl_stub_entry > .long sl_gdt > > will create relocations. The third one is easy, just replace it with > sl_gdt - sl_gdt_desc and initialize it at runtime with > > leal rva(sl_gdt_desc)(%ebx), %eax > addl %eax, 2(%eax) > lgdt (%eax) > > The other two are more messy, unfortunately there is no easy way to tell > the linker what we want here. The other entry point addresses (for the > EFI stub) are populated in a post-processing step after the compressed > kernel has been linked, we could teach it to also update kernel_info. > > Without that, for kernel_info, you could change it to store the offset > of the MLE header from kernel_info, instead of from the start of the > image. > > For the MLE header, it could be moved to .head.text in head_64.S, and > initialized with > .long rva(sl_stub) > This will also let it be placed at a fixed offset from startup_32, so > that kernel_info can just be populated with a constant. Thank you for the detailed reply. I am going to start digging into this now. Ross >
On Tue, Sep 29, 2020 at 10:03:47AM -0400, Ross Philipson wrote: > On 9/25/20 3:18 PM, Arvind Sankar wrote: > > You will also need to avoid initializing data with symbol addresses. > > > > .long mle_header > > .long sl_stub_entry > > .long sl_gdt > > ... > > > > The other two are more messy, unfortunately there is no easy way to tell > > the linker what we want here. The other entry point addresses (for the > > EFI stub) are populated in a post-processing step after the compressed > > kernel has been linked, we could teach it to also update kernel_info. > > > > Without that, for kernel_info, you could change it to store the offset > > of the MLE header from kernel_info, instead of from the start of the > > image. Actually, kernel_info is currently placed inside .rodata, which is quite a ways into the compressed kernel, so an offset from kernel_info would end up having to be negative, which would be ugly. I'll see if I can come up with some way to avoid this. > > > > For the MLE header, it could be moved to .head.text in head_64.S, and > > initialized with > > .long rva(sl_stub) > > This will also let it be placed at a fixed offset from startup_32, so > > that kernel_info can just be populated with a constant. > > Thank you for the detailed reply. I am going to start digging into this now. > > Ross > > > >
On Tue, Sep 29, 2020 at 10:03:47AM -0400, Ross Philipson wrote: > On 9/25/20 3:18 PM, Arvind Sankar wrote: [...] > > You should see them if you do > > readelf -r arch/x86/boot/compressed/vmlinux > > > > In terms of the code, things like: > > > > addl %ebx, (sl_gdt_desc + 2)(%ebx) > > > > will create a relocation, because the linker interprets this as wanting > > the runtime address of sl_gdt_desc, rather than just the offset from > > startup_32. > > > > https://urldefense.com/v3/__https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/tree/arch/x86/boot/compressed/head_64.S*n48__;Iw!!GqivPVa7Brio!JpZWv1cCPZdjD2jbCCGT7P9UIVl_lhX7YjckAnUcvi927jwZI7X3nX0MpIAZOyktJds$ > > > > has a comment with some explanation and a macro that the 32-bit code in > > startup_32 uses to avoid creating relocations. > > > > Since the SL code is in a different assembler file (and a different > > section), you can't directly use the same macro. I would suggest getting > > rid of sl_stub_entry and entering directly at sl_stub, and then the code > > in sl_stub.S can use sl_stub for the base address, defining the rva() > > macro there as > > > > #define rva(X) ((X) - sl_stub) > > > > You will also need to avoid initializing data with symbol addresses. > > > > .long mle_header > > .long sl_stub_entry > > .long sl_gdt > > > > will create relocations. The third one is easy, just replace it with > > sl_gdt - sl_gdt_desc and initialize it at runtime with > > > > leal rva(sl_gdt_desc)(%ebx), %eax > > addl %eax, 2(%eax) > > lgdt (%eax) > > > > The other two are more messy, unfortunately there is no easy way to tell > > the linker what we want here. The other entry point addresses (for the > > EFI stub) are populated in a post-processing step after the compressed > > kernel has been linked, we could teach it to also update kernel_info. > > > > Without that, for kernel_info, you could change it to store the offset > > of the MLE header from kernel_info, instead of from the start of the > > image. > > > > For the MLE header, it could be moved to .head.text in head_64.S, and > > initialized with > > .long rva(sl_stub) > > This will also let it be placed at a fixed offset from startup_32, so > > that kernel_info can just be populated with a constant. I am discussing with Ross the other option. We can create .rodata.mle_header section and put it at fixed offset as kernel_info is. So, we would have, e.g.: arch/x86/boot/compressed/vmlinux.lds.S: .rodata.kernel_info KERNEL_INFO_OFFSET : { *(.rodata.kernel_info) } ASSERT(ABSOLUTE(kernel_info) == KERNEL_INFO_OFFSET, "kernel_info at bad address!") .rodata.mle_header MLE_HEADER_OFFSET : { *(.rodata.mle_header) } ASSERT(ABSOLUTE(mle_header) == MLE_HEADER_OFFSET, "mle_header at bad address!") arch/x86/boot/compressed/sl_stub.S: #define mleh_rva(X) (((X) - mle_header) + MLE_HEADER_OFFSET) .section ".rodata.mle_header", "a" SYM_DATA_START(mle_header) .long 0x9082ac5a /* UUID0 */ .long 0x74a7476f /* UUID1 */ .long 0xa2555c0f /* UUID2 */ .long 0x42b651cb /* UUID3 */ .long 0x00000034 /* MLE header size */ .long 0x00020002 /* MLE version 2.2 */ .long mleh_rva(sl_stub_entry) /* Linear entry point of MLE (virt. address) */ .long 0x00000000 /* First valid page of MLE */ .long 0x00000000 /* Offset within binary of first byte of MLE */ .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ .long 0x00000223 /* Bit vector of MLE-supported capabilities */ .long 0x00000000 /* Starting linear address of command line (unused) */ .long 0x00000000 /* Ending linear address of command line (unused) */ SYM_DATA_END(mle_header) Of course MLE_HEADER_OFFSET has to be defined as a constant somewhere. Anyway, is it acceptable? There is also another problem. We have to put into mle_header size of the Linux kernel image. Currently it is done by the bootloader but I think it is not a role of the bootloader. The kernel image should provide all data describing its properties and do not rely on the bootloader to do that. Ross and I investigated various options but we did not find a good/simple way to do that. Could you suggest how we should do that or at least where we should take a look to get some ideas? Daniel
On Thu, Oct 15, 2020 at 08:26:54PM +0200, Daniel Kiper wrote: > > I am discussing with Ross the other option. We can create > .rodata.mle_header section and put it at fixed offset as > kernel_info is. So, we would have, e.g.: > > arch/x86/boot/compressed/vmlinux.lds.S: > .rodata.kernel_info KERNEL_INFO_OFFSET : { > *(.rodata.kernel_info) > } > ASSERT(ABSOLUTE(kernel_info) == KERNEL_INFO_OFFSET, "kernel_info at bad address!") > > .rodata.mle_header MLE_HEADER_OFFSET : { > *(.rodata.mle_header) > } > ASSERT(ABSOLUTE(mle_header) == MLE_HEADER_OFFSET, "mle_header at bad address!") > > arch/x86/boot/compressed/sl_stub.S: > #define mleh_rva(X) (((X) - mle_header) + MLE_HEADER_OFFSET) > > .section ".rodata.mle_header", "a" > > SYM_DATA_START(mle_header) > .long 0x9082ac5a /* UUID0 */ > .long 0x74a7476f /* UUID1 */ > .long 0xa2555c0f /* UUID2 */ > .long 0x42b651cb /* UUID3 */ > .long 0x00000034 /* MLE header size */ > .long 0x00020002 /* MLE version 2.2 */ > .long mleh_rva(sl_stub_entry) /* Linear entry point of MLE (virt. address) */ > .long 0x00000000 /* First valid page of MLE */ > .long 0x00000000 /* Offset within binary of first byte of MLE */ > .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ > .long 0x00000223 /* Bit vector of MLE-supported capabilities */ > .long 0x00000000 /* Starting linear address of command line (unused) */ > .long 0x00000000 /* Ending linear address of command line (unused) */ > SYM_DATA_END(mle_header) > > Of course MLE_HEADER_OFFSET has to be defined as a constant somewhere. > Anyway, is it acceptable? > > There is also another problem. We have to put into mle_header size of > the Linux kernel image. Currently it is done by the bootloader but > I think it is not a role of the bootloader. The kernel image should > provide all data describing its properties and do not rely on the > bootloader to do that. Ross and I investigated various options but we > did not find a good/simple way to do that. Could you suggest how we > should do that or at least where we should take a look to get some > ideas? > > Daniel What exactly is the size you need here? Is it just the size of the protected mode image, that's startup_32 to _edata. Or is it the size of the whole bzImage file, or something else? I guess the same question applies to "first valid page of MLE" and "first byte of MLE", and the linear entry point -- are those all relative to startup_32 or do they need to be relative to the start of the bzImage, i.e. you have to add the size of the real-mode boot stub? If you need to include the size of the bzImage file, that's not known when the files in boot/compressed are built. It's only known after the real-mode stub is linked. arch/x86/boot/tools/build.c fills in various details in the setup header and creates the bzImage file, but it does not currently modify anything in the protected-mode portion of the compressed kernel (i.e. arch/x86/boot/compressed/vmlinux, which then gets converted to binary format as arch/x86/boot/vmlinux.bin), so it would need to be extended if you need to modify the MLE header to include the bzImage size or anything depending on the size of the real-mode stub.
On 10/16/20 4:51 PM, Arvind Sankar wrote: > On Thu, Oct 15, 2020 at 08:26:54PM +0200, Daniel Kiper wrote: >> >> I am discussing with Ross the other option. We can create >> .rodata.mle_header section and put it at fixed offset as >> kernel_info is. So, we would have, e.g.: >> >> arch/x86/boot/compressed/vmlinux.lds.S: >> .rodata.kernel_info KERNEL_INFO_OFFSET : { >> *(.rodata.kernel_info) >> } >> ASSERT(ABSOLUTE(kernel_info) == KERNEL_INFO_OFFSET, "kernel_info at bad address!") >> >> .rodata.mle_header MLE_HEADER_OFFSET : { >> *(.rodata.mle_header) >> } >> ASSERT(ABSOLUTE(mle_header) == MLE_HEADER_OFFSET, "mle_header at bad address!") >> >> arch/x86/boot/compressed/sl_stub.S: >> #define mleh_rva(X) (((X) - mle_header) + MLE_HEADER_OFFSET) >> >> .section ".rodata.mle_header", "a" >> >> SYM_DATA_START(mle_header) >> .long 0x9082ac5a /* UUID0 */ >> .long 0x74a7476f /* UUID1 */ >> .long 0xa2555c0f /* UUID2 */ >> .long 0x42b651cb /* UUID3 */ >> .long 0x00000034 /* MLE header size */ >> .long 0x00020002 /* MLE version 2.2 */ >> .long mleh_rva(sl_stub_entry) /* Linear entry point of MLE (virt. address) */ >> .long 0x00000000 /* First valid page of MLE */ >> .long 0x00000000 /* Offset within binary of first byte of MLE */ >> .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ >> .long 0x00000223 /* Bit vector of MLE-supported capabilities */ >> .long 0x00000000 /* Starting linear address of command line (unused) */ >> .long 0x00000000 /* Ending linear address of command line (unused) */ >> SYM_DATA_END(mle_header) >> >> Of course MLE_HEADER_OFFSET has to be defined as a constant somewhere. >> Anyway, is it acceptable? >> >> There is also another problem. We have to put into mle_header size of >> the Linux kernel image. Currently it is done by the bootloader but >> I think it is not a role of the bootloader. The kernel image should >> provide all data describing its properties and do not rely on the >> bootloader to do that. Ross and I investigated various options but we >> did not find a good/simple way to do that. Could you suggest how we >> should do that or at least where we should take a look to get some >> ideas? >> >> Daniel > > What exactly is the size you need here? Is it just the size of the > protected mode image, that's startup_32 to _edata. Or is it the size of It is the size of the protected mode image. Though how to reference those symbols to get the size might all more relocation issues. > the whole bzImage file, or something else? I guess the same question > applies to "first valid page of MLE" and "first byte of MLE", and the Because of the way the launch environment is setup, both "first valid page of MLE" and "first byte of MLE" are always zero so nothing to have to sort out there. The only fields that need to be updated are "Linear entry point of MLE" and "Offset within binary of last byte + 1 of MLE". Thanks Ross > linear entry point -- are those all relative to startup_32 or do they > need to be relative to the start of the bzImage, i.e. you have to add > the size of the real-mode boot stub? > > If you need to include the size of the bzImage file, that's not known > when the files in boot/compressed are built. It's only known after the > real-mode stub is linked. arch/x86/boot/tools/build.c fills in various > details in the setup header and creates the bzImage file, but it does > not currently modify anything in the protected-mode portion of the > compressed kernel (i.e. arch/x86/boot/compressed/vmlinux, which then > gets converted to binary format as arch/x86/boot/vmlinux.bin), so it > would need to be extended if you need to modify the MLE header to > include the bzImage size or anything depending on the size of the > real-mode stub. >
On Fri, Oct 16, 2020 at 04:51:51PM -0400, Arvind Sankar wrote: > On Thu, Oct 15, 2020 at 08:26:54PM +0200, Daniel Kiper wrote: > > > > I am discussing with Ross the other option. We can create > > .rodata.mle_header section and put it at fixed offset as > > kernel_info is. So, we would have, e.g.: > > > > arch/x86/boot/compressed/vmlinux.lds.S: > > .rodata.kernel_info KERNEL_INFO_OFFSET : { > > *(.rodata.kernel_info) > > } > > ASSERT(ABSOLUTE(kernel_info) == KERNEL_INFO_OFFSET, "kernel_info at bad address!") > > > > .rodata.mle_header MLE_HEADER_OFFSET : { > > *(.rodata.mle_header) > > } > > ASSERT(ABSOLUTE(mle_header) == MLE_HEADER_OFFSET, "mle_header at bad address!") > > > > arch/x86/boot/compressed/sl_stub.S: > > #define mleh_rva(X) (((X) - mle_header) + MLE_HEADER_OFFSET) > > > > .section ".rodata.mle_header", "a" > > > > SYM_DATA_START(mle_header) > > .long 0x9082ac5a /* UUID0 */ > > .long 0x74a7476f /* UUID1 */ > > .long 0xa2555c0f /* UUID2 */ > > .long 0x42b651cb /* UUID3 */ > > .long 0x00000034 /* MLE header size */ > > .long 0x00020002 /* MLE version 2.2 */ > > .long mleh_rva(sl_stub_entry) /* Linear entry point of MLE (virt. address) */ > > .long 0x00000000 /* First valid page of MLE */ > > .long 0x00000000 /* Offset within binary of first byte of MLE */ > > .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ > > .long 0x00000223 /* Bit vector of MLE-supported capabilities */ > > .long 0x00000000 /* Starting linear address of command line (unused) */ > > .long 0x00000000 /* Ending linear address of command line (unused) */ > > SYM_DATA_END(mle_header) > > > > Of course MLE_HEADER_OFFSET has to be defined as a constant somewhere. > > Anyway, is it acceptable? What do you think about my MLE_HEADER_OFFSET and related stuff proposal? > > There is also another problem. We have to put into mle_header size of > > the Linux kernel image. Currently it is done by the bootloader but > > I think it is not a role of the bootloader. The kernel image should > > provide all data describing its properties and do not rely on the > > bootloader to do that. Ross and I investigated various options but we > > did not find a good/simple way to do that. Could you suggest how we > > should do that or at least where we should take a look to get some > > ideas? > > > > Daniel > > What exactly is the size you need here? Is it just the size of the > protected mode image, that's startup_32 to _edata. Or is it the size of > the whole bzImage file, or something else? I guess the same question > applies to "first valid page of MLE" and "first byte of MLE", and the > linear entry point -- are those all relative to startup_32 or do they > need to be relative to the start of the bzImage, i.e. you have to add > the size of the real-mode boot stub? > > If you need to include the size of the bzImage file, that's not known > when the files in boot/compressed are built. It's only known after the > real-mode stub is linked. arch/x86/boot/tools/build.c fills in various > details in the setup header and creates the bzImage file, but it does > not currently modify anything in the protected-mode portion of the > compressed kernel (i.e. arch/x86/boot/compressed/vmlinux, which then > gets converted to binary format as arch/x86/boot/vmlinux.bin), so it > would need to be extended if you need to modify the MLE header to > include the bzImage size or anything depending on the size of the > real-mode stub. Ross clarified this. So, I not have to add much here. Daniel
On Mon, Oct 19, 2020 at 10:38:08AM -0400, Ross Philipson wrote: > On 10/16/20 4:51 PM, Arvind Sankar wrote: > > On Thu, Oct 15, 2020 at 08:26:54PM +0200, Daniel Kiper wrote: > >> > >> I am discussing with Ross the other option. We can create > >> .rodata.mle_header section and put it at fixed offset as > >> kernel_info is. So, we would have, e.g.: > >> > >> arch/x86/boot/compressed/vmlinux.lds.S: > >> .rodata.kernel_info KERNEL_INFO_OFFSET : { > >> *(.rodata.kernel_info) > >> } > >> ASSERT(ABSOLUTE(kernel_info) == KERNEL_INFO_OFFSET, "kernel_info at bad address!") > >> > >> .rodata.mle_header MLE_HEADER_OFFSET : { > >> *(.rodata.mle_header) > >> } > >> ASSERT(ABSOLUTE(mle_header) == MLE_HEADER_OFFSET, "mle_header at bad address!") > >> > >> arch/x86/boot/compressed/sl_stub.S: > >> #define mleh_rva(X) (((X) - mle_header) + MLE_HEADER_OFFSET) > >> > >> .section ".rodata.mle_header", "a" > >> > >> SYM_DATA_START(mle_header) > >> .long 0x9082ac5a /* UUID0 */ > >> .long 0x74a7476f /* UUID1 */ > >> .long 0xa2555c0f /* UUID2 */ > >> .long 0x42b651cb /* UUID3 */ > >> .long 0x00000034 /* MLE header size */ > >> .long 0x00020002 /* MLE version 2.2 */ > >> .long mleh_rva(sl_stub_entry) /* Linear entry point of MLE (virt. address) */ > >> .long 0x00000000 /* First valid page of MLE */ > >> .long 0x00000000 /* Offset within binary of first byte of MLE */ > >> .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ > >> .long 0x00000223 /* Bit vector of MLE-supported capabilities */ > >> .long 0x00000000 /* Starting linear address of command line (unused) */ > >> .long 0x00000000 /* Ending linear address of command line (unused) */ > >> SYM_DATA_END(mle_header) > >> > >> Of course MLE_HEADER_OFFSET has to be defined as a constant somewhere. > >> Anyway, is it acceptable? > >> > >> There is also another problem. We have to put into mle_header size of > >> the Linux kernel image. Currently it is done by the bootloader but > >> I think it is not a role of the bootloader. The kernel image should > >> provide all data describing its properties and do not rely on the > >> bootloader to do that. Ross and I investigated various options but we > >> did not find a good/simple way to do that. Could you suggest how we > >> should do that or at least where we should take a look to get some > >> ideas? > >> > >> Daniel > > > > What exactly is the size you need here? Is it just the size of the > > protected mode image, that's startup_32 to _edata. Or is it the size of > > It is the size of the protected mode image. Though how to reference > those symbols to get the size might all more relocation issues. > Ok, then I think mleh_rva(_edata) should get you that -- I assume you don't want to include the uninitialized data in the size? The kernel will access memory beyond _edata (upto the init_size in the setup header), but that's not part of the image itself.
On Mon, Oct 19, 2020 at 04:51:53PM +0200, Daniel Kiper wrote: > On Fri, Oct 16, 2020 at 04:51:51PM -0400, Arvind Sankar wrote: > > On Thu, Oct 15, 2020 at 08:26:54PM +0200, Daniel Kiper wrote: > > > > > > I am discussing with Ross the other option. We can create > > > .rodata.mle_header section and put it at fixed offset as > > > kernel_info is. So, we would have, e.g.: > > > > > > arch/x86/boot/compressed/vmlinux.lds.S: > > > .rodata.kernel_info KERNEL_INFO_OFFSET : { > > > *(.rodata.kernel_info) > > > } > > > ASSERT(ABSOLUTE(kernel_info) == KERNEL_INFO_OFFSET, "kernel_info at bad address!") > > > > > > .rodata.mle_header MLE_HEADER_OFFSET : { > > > *(.rodata.mle_header) > > > } > > > ASSERT(ABSOLUTE(mle_header) == MLE_HEADER_OFFSET, "mle_header at bad address!") > > > > > > arch/x86/boot/compressed/sl_stub.S: > > > #define mleh_rva(X) (((X) - mle_header) + MLE_HEADER_OFFSET) > > > > > > .section ".rodata.mle_header", "a" > > > > > > SYM_DATA_START(mle_header) > > > .long 0x9082ac5a /* UUID0 */ > > > .long 0x74a7476f /* UUID1 */ > > > .long 0xa2555c0f /* UUID2 */ > > > .long 0x42b651cb /* UUID3 */ > > > .long 0x00000034 /* MLE header size */ > > > .long 0x00020002 /* MLE version 2.2 */ > > > .long mleh_rva(sl_stub_entry) /* Linear entry point of MLE (virt. address) */ > > > .long 0x00000000 /* First valid page of MLE */ > > > .long 0x00000000 /* Offset within binary of first byte of MLE */ > > > .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ > > > .long 0x00000223 /* Bit vector of MLE-supported capabilities */ > > > .long 0x00000000 /* Starting linear address of command line (unused) */ > > > .long 0x00000000 /* Ending linear address of command line (unused) */ > > > SYM_DATA_END(mle_header) > > > > > > Of course MLE_HEADER_OFFSET has to be defined as a constant somewhere. > > > Anyway, is it acceptable? > > What do you think about my MLE_HEADER_OFFSET and related stuff proposal? > I'm wondering if it would be easier to just allow relocations in these special "header" sections. I need to check how easy/hard it is to do that without triggering linker warnings. Thanks.
On 10/19/20 1:06 PM, Arvind Sankar wrote: > On Mon, Oct 19, 2020 at 10:38:08AM -0400, Ross Philipson wrote: >> On 10/16/20 4:51 PM, Arvind Sankar wrote: >>> On Thu, Oct 15, 2020 at 08:26:54PM +0200, Daniel Kiper wrote: >>>> >>>> I am discussing with Ross the other option. We can create >>>> .rodata.mle_header section and put it at fixed offset as >>>> kernel_info is. So, we would have, e.g.: >>>> >>>> arch/x86/boot/compressed/vmlinux.lds.S: >>>> .rodata.kernel_info KERNEL_INFO_OFFSET : { >>>> *(.rodata.kernel_info) >>>> } >>>> ASSERT(ABSOLUTE(kernel_info) == KERNEL_INFO_OFFSET, "kernel_info at bad address!") >>>> >>>> .rodata.mle_header MLE_HEADER_OFFSET : { >>>> *(.rodata.mle_header) >>>> } >>>> ASSERT(ABSOLUTE(mle_header) == MLE_HEADER_OFFSET, "mle_header at bad address!") >>>> >>>> arch/x86/boot/compressed/sl_stub.S: >>>> #define mleh_rva(X) (((X) - mle_header) + MLE_HEADER_OFFSET) >>>> >>>> .section ".rodata.mle_header", "a" >>>> >>>> SYM_DATA_START(mle_header) >>>> .long 0x9082ac5a /* UUID0 */ >>>> .long 0x74a7476f /* UUID1 */ >>>> .long 0xa2555c0f /* UUID2 */ >>>> .long 0x42b651cb /* UUID3 */ >>>> .long 0x00000034 /* MLE header size */ >>>> .long 0x00020002 /* MLE version 2.2 */ >>>> .long mleh_rva(sl_stub_entry) /* Linear entry point of MLE (virt. address) */ >>>> .long 0x00000000 /* First valid page of MLE */ >>>> .long 0x00000000 /* Offset within binary of first byte of MLE */ >>>> .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ >>>> .long 0x00000223 /* Bit vector of MLE-supported capabilities */ >>>> .long 0x00000000 /* Starting linear address of command line (unused) */ >>>> .long 0x00000000 /* Ending linear address of command line (unused) */ >>>> SYM_DATA_END(mle_header) >>>> >>>> Of course MLE_HEADER_OFFSET has to be defined as a constant somewhere. >>>> Anyway, is it acceptable? >>>> >>>> There is also another problem. We have to put into mle_header size of >>>> the Linux kernel image. Currently it is done by the bootloader but >>>> I think it is not a role of the bootloader. The kernel image should >>>> provide all data describing its properties and do not rely on the >>>> bootloader to do that. Ross and I investigated various options but we >>>> did not find a good/simple way to do that. Could you suggest how we >>>> should do that or at least where we should take a look to get some >>>> ideas? >>>> >>>> Daniel >>> >>> What exactly is the size you need here? Is it just the size of the >>> protected mode image, that's startup_32 to _edata. Or is it the size of >> >> It is the size of the protected mode image. Though how to reference >> those symbols to get the size might all more relocation issues. >> > > Ok, then I think mleh_rva(_edata) should get you that -- I assume you > don't want to include the uninitialized data in the size? The kernel > will access memory beyond _edata (upto the init_size in the setup > header), but that's not part of the image itself. Yea we basically want the size of the image. There is nothing to measure beyond the image as loaded into memory by the bootloader. rva(_edata) seems to be the ticket. Thanks Ross >
On Mon, Oct 19, 2020 at 01:18:22PM -0400, Arvind Sankar wrote: > On Mon, Oct 19, 2020 at 04:51:53PM +0200, Daniel Kiper wrote: > > On Fri, Oct 16, 2020 at 04:51:51PM -0400, Arvind Sankar wrote: > > > On Thu, Oct 15, 2020 at 08:26:54PM +0200, Daniel Kiper wrote: > > > > > > > > I am discussing with Ross the other option. We can create > > > > .rodata.mle_header section and put it at fixed offset as > > > > kernel_info is. So, we would have, e.g.: > > > > > > > > arch/x86/boot/compressed/vmlinux.lds.S: > > > > .rodata.kernel_info KERNEL_INFO_OFFSET : { > > > > *(.rodata.kernel_info) > > > > } > > > > ASSERT(ABSOLUTE(kernel_info) == KERNEL_INFO_OFFSET, "kernel_info at bad address!") > > > > > > > > .rodata.mle_header MLE_HEADER_OFFSET : { > > > > *(.rodata.mle_header) > > > > } > > > > ASSERT(ABSOLUTE(mle_header) == MLE_HEADER_OFFSET, "mle_header at bad address!") > > > > > > > > arch/x86/boot/compressed/sl_stub.S: > > > > #define mleh_rva(X) (((X) - mle_header) + MLE_HEADER_OFFSET) > > > > > > > > .section ".rodata.mle_header", "a" > > > > > > > > SYM_DATA_START(mle_header) > > > > .long 0x9082ac5a /* UUID0 */ > > > > .long 0x74a7476f /* UUID1 */ > > > > .long 0xa2555c0f /* UUID2 */ > > > > .long 0x42b651cb /* UUID3 */ > > > > .long 0x00000034 /* MLE header size */ > > > > .long 0x00020002 /* MLE version 2.2 */ > > > > .long mleh_rva(sl_stub_entry) /* Linear entry point of MLE (virt. address) */ > > > > .long 0x00000000 /* First valid page of MLE */ > > > > .long 0x00000000 /* Offset within binary of first byte of MLE */ > > > > .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ > > > > .long 0x00000223 /* Bit vector of MLE-supported capabilities */ > > > > .long 0x00000000 /* Starting linear address of command line (unused) */ > > > > .long 0x00000000 /* Ending linear address of command line (unused) */ > > > > SYM_DATA_END(mle_header) > > > > > > > > Of course MLE_HEADER_OFFSET has to be defined as a constant somewhere. > > > > Anyway, is it acceptable? > > > > What do you think about my MLE_HEADER_OFFSET and related stuff proposal? > > > > I'm wondering if it would be easier to just allow relocations in these > special "header" sections. I need to check how easy/hard it is to do > that without triggering linker warnings. Ross and I still bouncing some ideas. We came to the conclusion that putting mle_header into kernel .rodata.kernel_info section or even arch/x86/boot/compressed/kernel_info.S file would be the easiest thing to do at this point. Of course I would suggest some renaming too. E.g. .rodata.kernel_info to .rodata.kernel_headers, etc. Does it make sense for you? Daniel
On Wed, Oct 21, 2020 at 05:28:33PM +0200, Daniel Kiper wrote: > On Mon, Oct 19, 2020 at 01:18:22PM -0400, Arvind Sankar wrote: > > On Mon, Oct 19, 2020 at 04:51:53PM +0200, Daniel Kiper wrote: > > > On Fri, Oct 16, 2020 at 04:51:51PM -0400, Arvind Sankar wrote: > > > > On Thu, Oct 15, 2020 at 08:26:54PM +0200, Daniel Kiper wrote: > > > > > > > > > > I am discussing with Ross the other option. We can create > > > > > .rodata.mle_header section and put it at fixed offset as > > > > > kernel_info is. So, we would have, e.g.: > > > > > > > > > > arch/x86/boot/compressed/vmlinux.lds.S: > > > > > .rodata.kernel_info KERNEL_INFO_OFFSET : { > > > > > *(.rodata.kernel_info) > > > > > } > > > > > ASSERT(ABSOLUTE(kernel_info) == KERNEL_INFO_OFFSET, "kernel_info at bad address!") > > > > > > > > > > .rodata.mle_header MLE_HEADER_OFFSET : { > > > > > *(.rodata.mle_header) > > > > > } > > > > > ASSERT(ABSOLUTE(mle_header) == MLE_HEADER_OFFSET, "mle_header at bad address!") > > > > > > > > > > arch/x86/boot/compressed/sl_stub.S: > > > > > #define mleh_rva(X) (((X) - mle_header) + MLE_HEADER_OFFSET) > > > > > > > > > > .section ".rodata.mle_header", "a" > > > > > > > > > > SYM_DATA_START(mle_header) > > > > > .long 0x9082ac5a /* UUID0 */ > > > > > .long 0x74a7476f /* UUID1 */ > > > > > .long 0xa2555c0f /* UUID2 */ > > > > > .long 0x42b651cb /* UUID3 */ > > > > > .long 0x00000034 /* MLE header size */ > > > > > .long 0x00020002 /* MLE version 2.2 */ > > > > > .long mleh_rva(sl_stub_entry) /* Linear entry point of MLE (virt. address) */ > > > > > .long 0x00000000 /* First valid page of MLE */ > > > > > .long 0x00000000 /* Offset within binary of first byte of MLE */ > > > > > .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ > > > > > .long 0x00000223 /* Bit vector of MLE-supported capabilities */ > > > > > .long 0x00000000 /* Starting linear address of command line (unused) */ > > > > > .long 0x00000000 /* Ending linear address of command line (unused) */ > > > > > SYM_DATA_END(mle_header) > > > > > > > > > > Of course MLE_HEADER_OFFSET has to be defined as a constant somewhere. > > > > > Anyway, is it acceptable? > > > > > > What do you think about my MLE_HEADER_OFFSET and related stuff proposal? > > > > > > > I'm wondering if it would be easier to just allow relocations in these > > special "header" sections. I need to check how easy/hard it is to do > > that without triggering linker warnings. > > Ross and I still bouncing some ideas. We came to the conclusion that > putting mle_header into kernel .rodata.kernel_info section or even > arch/x86/boot/compressed/kernel_info.S file would be the easiest thing > to do at this point. Of course I would suggest some renaming too. E.g. > .rodata.kernel_info to .rodata.kernel_headers, etc. Does it make sense > for you? > > Daniel I haven't been able to come up with any different options that don't require post-processing of the kernel image. Allowing relocations in specific sections seems to not be possible with lld, and anyway would require the fields to be 64-bit sized so it doesn't really help. Putting mle_header into kernel_info seems like a reasonable thing to me, and if you do that, putting it into kernel_info.S would seem to be necessary? Would you also have a fixed field with the offset of the mle_header from kernel_info? That seems nicer than having the bootloader scan the variable data for magic strings.
On 10/21/20 12:18 PM, Arvind Sankar wrote: > On Wed, Oct 21, 2020 at 05:28:33PM +0200, Daniel Kiper wrote: >> On Mon, Oct 19, 2020 at 01:18:22PM -0400, Arvind Sankar wrote: >>> On Mon, Oct 19, 2020 at 04:51:53PM +0200, Daniel Kiper wrote: >>>> On Fri, Oct 16, 2020 at 04:51:51PM -0400, Arvind Sankar wrote: >>>>> On Thu, Oct 15, 2020 at 08:26:54PM +0200, Daniel Kiper wrote: >>>>>> >>>>>> I am discussing with Ross the other option. We can create >>>>>> .rodata.mle_header section and put it at fixed offset as >>>>>> kernel_info is. So, we would have, e.g.: >>>>>> >>>>>> arch/x86/boot/compressed/vmlinux.lds.S: >>>>>> .rodata.kernel_info KERNEL_INFO_OFFSET : { >>>>>> *(.rodata.kernel_info) >>>>>> } >>>>>> ASSERT(ABSOLUTE(kernel_info) == KERNEL_INFO_OFFSET, "kernel_info at bad address!") >>>>>> >>>>>> .rodata.mle_header MLE_HEADER_OFFSET : { >>>>>> *(.rodata.mle_header) >>>>>> } >>>>>> ASSERT(ABSOLUTE(mle_header) == MLE_HEADER_OFFSET, "mle_header at bad address!") >>>>>> >>>>>> arch/x86/boot/compressed/sl_stub.S: >>>>>> #define mleh_rva(X) (((X) - mle_header) + MLE_HEADER_OFFSET) >>>>>> >>>>>> .section ".rodata.mle_header", "a" >>>>>> >>>>>> SYM_DATA_START(mle_header) >>>>>> .long 0x9082ac5a /* UUID0 */ >>>>>> .long 0x74a7476f /* UUID1 */ >>>>>> .long 0xa2555c0f /* UUID2 */ >>>>>> .long 0x42b651cb /* UUID3 */ >>>>>> .long 0x00000034 /* MLE header size */ >>>>>> .long 0x00020002 /* MLE version 2.2 */ >>>>>> .long mleh_rva(sl_stub_entry) /* Linear entry point of MLE (virt. address) */ >>>>>> .long 0x00000000 /* First valid page of MLE */ >>>>>> .long 0x00000000 /* Offset within binary of first byte of MLE */ >>>>>> .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ >>>>>> .long 0x00000223 /* Bit vector of MLE-supported capabilities */ >>>>>> .long 0x00000000 /* Starting linear address of command line (unused) */ >>>>>> .long 0x00000000 /* Ending linear address of command line (unused) */ >>>>>> SYM_DATA_END(mle_header) >>>>>> >>>>>> Of course MLE_HEADER_OFFSET has to be defined as a constant somewhere. >>>>>> Anyway, is it acceptable? >>>> >>>> What do you think about my MLE_HEADER_OFFSET and related stuff proposal? >>>> >>> >>> I'm wondering if it would be easier to just allow relocations in these >>> special "header" sections. I need to check how easy/hard it is to do >>> that without triggering linker warnings. >> >> Ross and I still bouncing some ideas. We came to the conclusion that >> putting mle_header into kernel .rodata.kernel_info section or even >> arch/x86/boot/compressed/kernel_info.S file would be the easiest thing >> to do at this point. Of course I would suggest some renaming too. E.g. >> .rodata.kernel_info to .rodata.kernel_headers, etc. Does it make sense >> for you? >> >> Daniel > > I haven't been able to come up with any different options that don't > require post-processing of the kernel image. Allowing relocations in > specific sections seems to not be possible with lld, and anyway would > require the fields to be 64-bit sized so it doesn't really help. > > Putting mle_header into kernel_info seems like a reasonable thing to me, > and if you do that, putting it into kernel_info.S would seem to be > necessary? Would you also have a fixed field with the offset of the That seems like a reasonable place for it to go. > mle_header from kernel_info? That seems nicer than having the > bootloader scan the variable data for magic strings. Yes kernel_info will have a field to the offset of the mle_header. I agree that would be nicer. Thanks Ross >
diff --git a/Documentation/x86/boot.rst b/Documentation/x86/boot.rst index 7fafc7a..7232801 100644 --- a/Documentation/x86/boot.rst +++ b/Documentation/x86/boot.rst @@ -1026,6 +1026,15 @@ Offset/size: 0x000c/4 This field contains maximal allowed type for setup_data and setup_indirect structs. +============ ================= +Field name: mle_header_offset +Offset/size: 0x0010/4 +============ ================= + + This field contains the offset to the Secure Launch Measured Launch Environment + (MLE) header. This offset is used to locate information needed during a secure + late launch using Intel TXT and AMD SKINIT. + The Image Checksum ================== diff --git a/arch/x86/boot/compressed/Makefile b/arch/x86/boot/compressed/Makefile index 35947b9..d881ff7 100644 --- a/arch/x86/boot/compressed/Makefile +++ b/arch/x86/boot/compressed/Makefile @@ -102,6 +102,7 @@ vmlinux-objs-$(CONFIG_SECURE_LAUNCH_SHA512) += $(obj)/early_sha512.o vmlinux-objs-$(CONFIG_SECURE_LAUNCH) += $(obj)/tpm/tpmio.o $(obj)/tpm/tpm_buff.o \ $(obj)/tpm/tis.o $(obj)/tpm/crb.o $(obj)/tpm/tpm1_cmds.o \ $(obj)/tpm/tpm2_cmds.o $(obj)/tpm/tpm2_auth.o $(obj)/tpm/tpm.o +vmlinux-objs-$(CONFIG_SECURE_LAUNCH) += $(obj)/sl_main.o $(obj)/sl_stub.o # The compressed kernel is built with -fPIC/-fPIE so that a boot loader # can place it anywhere in memory and it will still run. However, since diff --git a/arch/x86/boot/compressed/head_64.S b/arch/x86/boot/compressed/head_64.S index 97d37f0..42043bf 100644 --- a/arch/x86/boot/compressed/head_64.S +++ b/arch/x86/boot/compressed/head_64.S @@ -279,6 +279,21 @@ SYM_INNER_LABEL(efi32_pe_stub_entry, SYM_L_LOCAL) SYM_FUNC_END(efi32_stub_entry) #endif +#ifdef CONFIG_SECURE_LAUNCH +SYM_FUNC_START(sl_stub_entry) + /* + * On entry, %ebx has the entry abs offset to sl_stub_entry. To + * find the beginning of where we are loaded, sub off from the + * beginning. + */ + leal (startup_32 - sl_stub_entry)(%ebx), %ebx + + /* More room to work in sl_stub in the text section */ + jmp sl_stub + +SYM_FUNC_END(sl_stub_entry) +#endif + .code64 .org 0x200 SYM_CODE_START(startup_64) @@ -537,6 +552,25 @@ SYM_FUNC_START_LOCAL_NOALIGN(.Lrelocated) shrq $3, %rcx rep stosq +#ifdef CONFIG_SECURE_LAUNCH + /* + * Have to do the final early sl stub work in 64b area. + * + * *********** NOTE *********** + * + * Several boot params get used before we get a chance to measure + * them in this call. This is a known issue and we currently don't + * have a solution. The scratch field doesn't matter and loadflags + * have KEEP_SEGMENTS set by the stub code. There is no obvious way + * to do anything about the use of kernel_alignment or init_size + * though these seem low risk. + */ + pushq %rsi + movq %rsi, %rdi + callq sl_main + popq %rsi +#endif + /* * Do the extraction, and jump to the new kernel.. */ diff --git a/arch/x86/boot/compressed/kernel_info.S b/arch/x86/boot/compressed/kernel_info.S index f818ee8..192d557 100644 --- a/arch/x86/boot/compressed/kernel_info.S +++ b/arch/x86/boot/compressed/kernel_info.S @@ -17,6 +17,13 @@ kernel_info: /* Maximal allowed type for setup_data and setup_indirect structs. */ .long SETUP_TYPE_MAX + /* Offset to the MLE header structure */ +#ifdef CONFIG_SECURE_LAUNCH + .long mle_header +#else + .long 0 +#endif + kernel_info_var_len_data: /* Empty for time being... */ kernel_info_end: diff --git a/arch/x86/boot/compressed/sl_main.c b/arch/x86/boot/compressed/sl_main.c new file mode 100644 index 0000000..09ec92d --- /dev/null +++ b/arch/x86/boot/compressed/sl_main.c @@ -0,0 +1,390 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Secure Launch early measurement and validation routines. + * + * Copyright (c) 2020, Oracle and/or its affiliates. + */ + +#include <linux/init.h> +#include <linux/string.h> +#include <linux/linkage.h> +#include <linux/efi.h> +#include <asm/segment.h> +#include <asm/boot.h> +#include <asm/msr.h> +#include <asm/io.h> +#include <asm/mtrr.h> +#include <asm/processor-flags.h> +#include <asm/asm-offsets.h> +#include <asm/bootparam.h> +#include <asm/efi.h> +#include <linux/slaunch.h> +#ifdef CONFIG_SECURE_LAUNCH_SHA256 +#include <config/crypto/sha256.h> +#endif +#ifdef CONFIG_SECURE_LAUNCH_SHA512 +#include <linux/sha512.h> +#endif + +#include "early_sha1.h" +#include "tpm/tpm_common.h" +#include "tpm/tpm2_constants.h" +#include "tpm/tpm.h" + +#define CAPS_VARIABLE_MTRR_COUNT_MASK 0xff + +#define SL_MAX_EVENT_DATA 64 +#define SL_TPM12_LOG_SIZE (sizeof(struct tpm12_pcr_event) + \ + SL_MAX_EVENT_DATA) +#define SL_TPM20_LOG_SIZE (sizeof(struct tpm20_ha) + \ + SHA512_SIZE + \ + sizeof(struct tpm20_digest_values) + \ + sizeof(struct tpm20_pcr_event_head) + \ + sizeof(struct tpm20_pcr_event_tail) + \ + SL_MAX_EVENT_DATA) + +static void *evtlog_base; +static struct txt_heap_event_log_pointer2_1_element *log20_elem; + +extern u32 sl_cpu_type; + +static u64 sl_txt_read(u32 reg) +{ + return readq((void *)(u64)(TXT_PRIV_CONFIG_REGS_BASE + reg)); +} + +static void sl_txt_write(u32 reg, u64 val) +{ + writeq(val, (void *)(u64)(TXT_PRIV_CONFIG_REGS_BASE + reg)); +} + +static void __noreturn sl_txt_reset(u64 error) +{ + /* Reading the E2STS register acts as a barrier for TXT registers */ + sl_txt_write(TXT_CR_ERRORCODE, error); + sl_txt_read(TXT_CR_E2STS); + sl_txt_write(TXT_CR_CMD_UNLOCK_MEM_CONFIG, 1); + sl_txt_read(TXT_CR_E2STS); + sl_txt_write(TXT_CR_CMD_RESET, 1); + + for ( ; ; ) + asm volatile ("hlt"); + + unreachable(); +} + +static u64 sl_rdmsr(u32 reg) +{ + u64 lo, hi; + + asm volatile ("rdmsr" : "=a" (lo), "=d" (hi) : "c" (reg)); + + return (hi << 32) | lo; +} + +/* + * Some MSRs are modified by the pre-launch code including the MTRRs. + * The early MLE code has to restore these values. This code validates + * the values after they are measured. + */ +static void sl_txt_validate_msrs(struct txt_os_mle_data *os_mle_data) +{ + u64 mtrr_caps, mtrr_def_type, mtrr_var, misc_en_msr; + u32 vcnt, i; + struct txt_mtrr_state *saved_bsp_mtrrs = + &(os_mle_data->saved_bsp_mtrrs); + + mtrr_caps = sl_rdmsr(MSR_MTRRcap); + vcnt = (u32)(mtrr_caps & CAPS_VARIABLE_MTRR_COUNT_MASK); + + if (saved_bsp_mtrrs->mtrr_vcnt > vcnt) + sl_txt_reset(SL_ERROR_MTRR_INV_VCNT); + if (saved_bsp_mtrrs->mtrr_vcnt > TXT_OS_MLE_MAX_VARIABLE_MTRRS) + sl_txt_reset(SL_ERROR_MTRR_INV_VCNT); + + mtrr_def_type = sl_rdmsr(MSR_MTRRdefType); + if (saved_bsp_mtrrs->default_mem_type != mtrr_def_type) + sl_txt_reset(SL_ERROR_MTRR_INV_DEF_TYPE); + + for (i = 0; i < saved_bsp_mtrrs->mtrr_vcnt; i++) { + mtrr_var = sl_rdmsr(MTRRphysBase_MSR(i)); + if (saved_bsp_mtrrs->mtrr_pair[i].mtrr_physbase != mtrr_var) + sl_txt_reset(SL_ERROR_MTRR_INV_BASE); + mtrr_var = sl_rdmsr(MTRRphysMask_MSR(i)); + if (saved_bsp_mtrrs->mtrr_pair[i].mtrr_physmask != mtrr_var) + sl_txt_reset(SL_ERROR_MTRR_INV_MASK); + } + + misc_en_msr = sl_rdmsr(MSR_IA32_MISC_ENABLE); + if (os_mle_data->saved_misc_enable_msr != misc_en_msr) + sl_txt_reset(SL_ERROR_MSR_INV_MISC_EN); +} + +static void sl_find_event_log(struct tpm *tpm) +{ + struct txt_os_mle_data *os_mle_data; + void *os_sinit_data; + void *txt_heap; + + txt_heap = (void *)sl_txt_read(TXT_CR_HEAP_BASE); + + os_mle_data = txt_os_mle_data_start(txt_heap); + evtlog_base = (void *)os_mle_data->evtlog_addr; + + if (tpm->family != TPM20) + return; + + /* + * For TPM 2.0, the event log 2.1 extended data structure has to also + * be located and fixed up. + */ + os_sinit_data = txt_os_sinit_data_start(txt_heap); + + /* Find the TPM2.0 logging extended heap element */ + log20_elem = tpm20_find_log2_1_element(os_sinit_data); + + if (!log20_elem) + sl_txt_reset(SL_ERROR_TPM_INVALID_LOG20); +} + +static void sl_tpm12_log_event(u32 pcr, u8 *digest, + const u8 *event_data, u32 event_size) +{ + struct tpm12_pcr_event *pcr_event; + u32 total_size; + u8 log_buf[SL_TPM12_LOG_SIZE] = {0}; + + pcr_event = (struct tpm12_pcr_event *)log_buf; + pcr_event->pcr_index = pcr; + pcr_event->type = TXT_EVTYPE_SLAUNCH; + memcpy(&pcr_event->digest[0], digest, SHA1_SIZE); + pcr_event->size = event_size; + memcpy((u8 *)pcr_event + sizeof(struct tpm12_pcr_event), + event_data, event_size); + + total_size = sizeof(struct tpm12_pcr_event) + event_size; + + if (tpm12_log_event(evtlog_base, total_size, pcr_event)) + sl_txt_reset(SL_ERROR_TPM_LOGGING_FAILED); +} + +static void sl_tpm20_log_event(u32 pcr, u8 *digest, u16 algo, + const u8 *event_data, u32 event_size) +{ + struct tpm20_pcr_event_head *head; + struct tpm20_digest_values *dvs; + struct tpm20_ha *ha; + struct tpm20_pcr_event_tail *tail; + u8 *dptr; + u32 total_size; + u8 log_buf[SL_TPM20_LOG_SIZE] = {0}; + + head = (struct tpm20_pcr_event_head *)log_buf; + head->pcr_index = pcr; + head->event_type = TXT_EVTYPE_SLAUNCH; + dvs = (struct tpm20_digest_values *) + ((u8 *)head + sizeof(struct tpm20_pcr_event_head)); + dvs->count = 1; + ha = (struct tpm20_ha *) + ((u8 *)dvs + sizeof(struct tpm20_digest_values)); + ha->algorithm_id = algo; + dptr = (u8 *)ha + sizeof(struct tpm20_ha); + + switch (algo) { + case TPM_ALG_SHA512: + memcpy(dptr, digest, SHA512_SIZE); + tail = (struct tpm20_pcr_event_tail *) + (dptr + SHA512_SIZE); + break; + case TPM_ALG_SHA256: + memcpy(dptr, digest, SHA256_SIZE); + tail = (struct tpm20_pcr_event_tail *) + (dptr + SHA256_SIZE); + break; + case TPM_ALG_SHA1: + default: + memcpy(dptr, digest, SHA1_SIZE); + tail = (struct tpm20_pcr_event_tail *) + (dptr + SHA1_SIZE); + }; + + tail->event_size = event_size; + memcpy((u8 *)tail + sizeof(struct tpm20_pcr_event_tail), + event_data, event_size); + + total_size = (u32)((u8 *)tail - (u8 *)head) + + sizeof(struct tpm20_pcr_event_tail) + event_size; + + if (tpm20_log_event(log20_elem, evtlog_base, total_size, &log_buf[0])) + sl_txt_reset(SL_ERROR_TPM_LOGGING_FAILED); +} + +void sl_tpm_extend_pcr(struct tpm *tpm, u32 pcr, const u8 *data, u32 length, + const char *desc) +{ + struct sha1_state sctx = {0}; + u8 sha1_hash[SHA1_SIZE] = {0}; + int ret; + + if (tpm->family == TPM20) { +#ifdef CONFIG_SECURE_LAUNCH_SHA256 + struct sha256_state sctx = {0}; + u8 sha256_hash[SHA256_SIZE] = {0}; + + sha256_init(&sctx); + sha256_update(&sctx, data, length); + sha256_final(&sctx, &sha256_hash[0]); + ret = tpm_extend_pcr(tpm, pcr, TPM_ALG_SHA256, &sha256_hash[0]); + if (!ret) { + sl_tpm20_log_event(pcr, &sha256_hash[0], + TPM_ALG_SHA256, + (const u8 *)desc, strlen(desc)); + return; + } else + sl_txt_reset(SL_ERROR_TPM_EXTEND); +#endif +#ifdef CONFIG_SECURE_LAUNCH_SHA512 + struct sha512_state sctx = {0}; + u8 sha512_hash[SHA512_SIZE] = {0}; + + sha512_init(&sctx); + sha512_update(&sctx, data, length); + sha512_final(&sctx, &sha512_hash[0]); + ret = tpm_extend_pcr(tpm, pcr, TPM_ALG_SHA512, &sha512_hash[0]); + if (!ret) { + sl_tpm20_log_event(pcr, &sha512_hash[0], + TPM_ALG_SHA512, + (const u8 *)desc, strlen(desc)); + return; + } else + sl_txt_reset(SL_ERROR_TPM_EXTEND); +#endif + } + + early_sha1_init(&sctx); + early_sha1_update(&sctx, data, length); + early_sha1_final(&sctx, &sha1_hash[0]); + ret = tpm_extend_pcr(tpm, pcr, TPM_ALG_SHA1, &sha1_hash[0]); + if (ret) + sl_txt_reset(SL_ERROR_TPM_EXTEND); + + if (tpm->family == TPM20) + sl_tpm20_log_event(pcr, &sha1_hash[0], TPM_ALG_SHA1, + (const u8 *)desc, strlen(desc)); + else + sl_tpm12_log_event(pcr, &sha1_hash[0], + (const u8 *)desc, strlen(desc)); +} + +void sl_main(u8 *bootparams) +{ + struct tpm *tpm; + struct boot_params *bp; + struct setup_data *data; + struct txt_os_mle_data *os_mle_data; + struct txt_os_mle_data os_mle_tmp = {0}; + const char *signature; + unsigned long mmap = 0; + void *txt_heap; + u32 data_count; + + /* + * Currently only Intel TXT is supported for Secure Launch. Testing + * this value also indicates that the kernel was booted successfully + * through the Secure Launch entry point and is in SMX mode. + */ + if (!(sl_cpu_type & SL_CPU_INTEL)) + return; + + /* + * If enable_tpm fails there is no point going on. The entire secure + * environment depends on this and the other TPM operations succeeding. + */ + tpm = enable_tpm(); + if (!tpm) + sl_txt_reset(SL_ERROR_TPM_INIT); + + /* Locate the TPM event log. */ + sl_find_event_log(tpm); + + /* + * Locality 2 is being opened so that the DRTM PCRs can be updated, + * specifically 17 and 18. + */ + if (tpm_request_locality(tpm, 2) == TPM_NO_LOCALITY) + sl_txt_reset(SL_ERROR_TPM_GET_LOC); + + /* Measure the zero page/boot params */ + sl_tpm_extend_pcr(tpm, SL_CONFIG_PCR18, bootparams, PAGE_SIZE, + "Measured boot parameters into PCR18"); + + /* Now safe to use boot params */ + bp = (struct boot_params *)bootparams; + + /* Measure the command line */ + sl_tpm_extend_pcr(tpm, SL_CONFIG_PCR18, + (u8 *)((unsigned long)bp->hdr.cmd_line_ptr), + bp->hdr.cmdline_size, + "Measured Kernel command line into PCR18"); + + /* + * Measuring the boot params measured the fixed e820 memory map. + * Measure any setup_data entries including e820 extended entries. + */ + data = (struct setup_data *)(unsigned long)bp->hdr.setup_data; + while (data) { + sl_tpm_extend_pcr(tpm, SL_CONFIG_PCR18, + ((u8 *)data) + sizeof(struct setup_data), + data->len, + "Measured Kernel setup_data into PCR18"); + + data = (struct setup_data *)(unsigned long)data->next; + } + + /* If bootloader was EFI, measure the memory map passed across */ + signature = + (const char *)&bp->efi_info.efi_loader_signature; + + if (!strncmp(signature, EFI32_LOADER_SIGNATURE, 4)) + mmap = bp->efi_info.efi_memmap; + else if (!strncmp(signature, EFI64_LOADER_SIGNATURE, 4)) + mmap = (bp->efi_info.efi_memmap | + ((u64)bp->efi_info.efi_memmap_hi << 32)); + + if (mmap) + sl_tpm_extend_pcr(tpm, SL_CONFIG_PCR18, (void *)mmap, + bp->efi_info.efi_memmap_size, + "Measured EFI memory map into PCR18"); + + /* Measure any external initrd */ + if (bp->hdr.ramdisk_image != 0 && bp->hdr.ramdisk_size != 0) + sl_tpm_extend_pcr(tpm, SL_IMAGE_PCR17, + (u8 *)((u64)bp->hdr.ramdisk_image), + bp->hdr.ramdisk_size, + "Measured initramfs into PCR17"); + + /* + * Some extra work to do on Intel, have to measure the OS-MLE + * heap area. + */ + txt_heap = (void *)sl_txt_read(TXT_CR_HEAP_BASE); + os_mle_data = txt_os_mle_data_start(txt_heap); + + /* Measure only portions of OS-MLE data, not addresses/sizes etc. */ + os_mle_tmp.version = os_mle_data->version; + os_mle_tmp.saved_misc_enable_msr = os_mle_data->saved_misc_enable_msr; + os_mle_tmp.saved_bsp_mtrrs = os_mle_data->saved_bsp_mtrrs; + + sl_tpm_extend_pcr(tpm, SL_CONFIG_PCR18, (u8 *)&os_mle_tmp, + sizeof(struct txt_os_mle_data), + "Measured TXT OS-MLE data into PCR18"); + + /* + * Now that the OS-MLE data is measured, ensure the MTRR and + * misc enable MSRs are what we expect. + */ + sl_txt_validate_msrs(os_mle_data); + + tpm_relinquish_locality(tpm); + free_tpm(tpm); +} diff --git a/arch/x86/boot/compressed/sl_stub.S b/arch/x86/boot/compressed/sl_stub.S new file mode 100644 index 0000000..a1e7c13 --- /dev/null +++ b/arch/x86/boot/compressed/sl_stub.S @@ -0,0 +1,606 @@ +/* SPDX-License-Identifier: GPL-2.0 */ + +/* + * Secure Launch protected mode entry point. + * + * Copyright (c) 2020, Oracle and/or its affiliates. + */ + .code32 + .text +#include <linux/linkage.h> +#include <asm/segment.h> +#include <asm/msr.h> +#include <asm/processor-flags.h> +#include <asm/asm-offsets.h> +#include <asm/bootparam.h> +#include <asm/irq_vectors.h> +#include <linux/slaunch.h> + +/* Can't include apiddef.h in asm */ +#define XAPIC_ENABLE (1 << 11) +#define X2APIC_ENABLE (1 << 10) + +/* Can't include traps.h in asm */ +#define X86_TRAP_NMI 2 + +/* Can't include mtrr.h in asm */ +#define MTRRphysBase0 0x200 + +#define IDT_VECTOR_LO_BITS 0 +#define IDT_VECTOR_HI_BITS 6 + +/* + * The GETSEC op code is open coded because older versions of + * GCC do not support the getsec mnemonic. + */ +.macro GETSEC leaf + pushl %ebx + xorl %ebx, %ebx /* Must be zero for SMCTRL */ + movl \leaf, %eax /* Leaf function */ + .byte 0x0f, 0x37 /* GETSEC opcode */ + popl %ebx +.endm + +.macro TXT_RESET error + /* + * Set a sticky error value and reset. Note the movs to %eax act as + * TXT register barriers. + */ + movl \error, (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_ERRORCODE) + movl (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_E2STS), %eax + movl $1, (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_CMD_NO_SECRETS) + movl (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_E2STS), %eax + movl $1, (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_CMD_UNLOCK_MEM_CONFIG) + movl (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_E2STS), %eax + movl $1, (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_CMD_RESET) +1: + hlt + jmp 1b +.endm + + /* + * The MLE Header per the TXT Specification, section 2.1 + * MLE capabilities, see table 4. Capabilities set: + * bit 0: Support for GETSEC[WAKEUP] for RLP wakeup + * bit 1: Support for RLP wakeup using MONITOR address + * bit 5: TPM 1.2 family: Details/authorities PCR usage support + * bit 9: Supported format of TPM 2.0 event log - TCG compliant + */ +SYM_DATA_START(mle_header) + .long 0x9082ac5a /* UUID0 */ + .long 0x74a7476f /* UUID1 */ + .long 0xa2555c0f /* UUID2 */ + .long 0x42b651cb /* UUID3 */ + .long 0x00000034 /* MLE header size */ + .long 0x00020002 /* MLE version 2.2 */ + .long sl_stub_entry /* Linear entry point of MLE (virt. address) */ + .long 0x00000000 /* First valid page of MLE */ + .long 0x00000000 /* Offset within binary of first byte of MLE */ + .long 0x00000000 /* Offset within binary of last byte + 1 of MLE */ + .long 0x00000223 /* Bit vector of MLE-supported capabilities */ + .long 0x00000000 /* Starting linear address of command line (unused) */ + .long 0x00000000 /* Ending linear address of command line (unused) */ +SYM_DATA_END(mle_header) + + .code32 +SYM_FUNC_START(sl_stub) + /* + * On entry, %ebx has the base address from head_64.S + * and only %cs and %ds segments are known good. + */ + cli + cld + + /* Load GDT, set segment regs and lret to __SL32_CS */ + addl %ebx, (sl_gdt_desc + 2)(%ebx) + lgdt sl_gdt_desc(%ebx) + + movl $(__SL32_DS), %eax + movw %ax, %ds + movw %ax, %es + movw %ax, %fs + movw %ax, %gs + movw %ax, %ss + + /* + * Now that %ss us known good, take the first stack for the BSP. The + * AP stacks are only used on Intel. + */ + leal sl_stacks_end(%ebx), %esp + + leal .Lsl_cs(%ebx), %eax + pushl $(__SL32_CS) + pushl %eax + lret + +.Lsl_cs: + /* Save our base pointer reg */ + pushl %ebx + + /* Now see if it is GenuineIntel. CPUID 0 returns the manufacturer */ + xorl %eax, %eax + cpuid + cmpl $(INTEL_CPUID_MFGID_EBX), %ebx + jnz .Ldo_unknown_cpu + cmpl $(INTEL_CPUID_MFGID_EDX), %edx + jnz .Ldo_unknown_cpu + cmpl $(INTEL_CPUID_MFGID_ECX), %ecx + jnz .Ldo_unknown_cpu + + popl %ebx + + /* Know it is Intel */ + movl $(SL_CPU_INTEL), sl_cpu_type(%ebx) + + /* Increment CPU count for BSP */ + incl sl_txt_cpu_count(%ebx) + + /* Enable SMI with GETSEC[SMCTRL] */ + GETSEC $(SMX_X86_GETSEC_SMCTRL) + + /* IRET-to-self can be used to enable NMIs which SENTER disabled */ + leal .Lnmi_enabled(%ebx), %eax + pushfl + pushl $(__SL32_CS) + pushl %eax + iret + +.Lnmi_enabled: + /* Clear the TXT error registers for a clean start of day */ + movl $0, (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_ERRORCODE) + movl $0xffffffff, (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_ESTS) + + /* On Intel, the zero page address is passed in the TXT heap */ + /* Read physical base of heap into EAX */ + movl (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_HEAP_BASE), %eax + /* Read the size of the BIOS data into ECX (first 8 bytes) */ + movl (%eax), %ecx + /* Skip over BIOS data and size of OS to MLE data section */ + leal 8(%eax, %ecx), %eax + + /* Check that the AP wake block is big enough */ + cmpl $(sl_txt_ap_wake_end - sl_txt_ap_wake_begin), \ + SL_ap_wake_block_size(%eax) + jae .Lwake_block_ok + TXT_RESET $(SL_ERROR_WAKE_BLOCK_TOO_SMALL) + +.Lwake_block_ok: + /* + * Get the boot params address from the heap. Note %esi and %ebx MUST + * be preserved across calls and operations. + */ + movl SL_boot_params_addr(%eax), %esi + + /* Save %ebx so the APs can find their way home */ + movl %ebx, (SL_mle_scratch + SL_SCRATCH_AP_EBX)(%eax) + + /* Fetch the AP wake code block address from the heap */ + movl SL_ap_wake_block(%eax), %edi + movl %edi, sl_txt_ap_wake_block(%ebx) + + /* Store the offset in the AP wake block to the jmp address */ + movl $(sl_ap_jmp_offset - sl_txt_ap_wake_begin), \ + (SL_mle_scratch + SL_SCRATCH_AP_JMP_OFFSET)(%eax) + + /* %eax still is the base of the OS-MLE block, save it */ + pushl %eax + + /* Relocate the AP wake code to the safe block */ + call sl_txt_reloc_ap_wake + + /* + * Wake up all APs that are blocked in the ACM and wait for them to + * halt. This should be done before restoring the MTRRs so the ACM is + * still properly in WB memory. + */ + call sl_txt_wake_aps + + /* + * Pop OS-MLE base address (was in %eax above) for call to load + * MTRRs/MISC MSR + */ + popl %edi + call sl_txt_load_regs + + jmp .Lcpu_setup_done + +.Ldo_unknown_cpu: + /* Non-Intel CPUs are not yet supported */ + ud2 + +.Lcpu_setup_done: + /* + * Don't enable MCE at this point. The kernel will enable + * it on the BSP later when it is ready. + */ + + /* Keep SL segments for the early portion of the kernel boot */ + orb $(KEEP_SEGMENTS), BP_loadflags(%esi) + + /* Done, jump to normal 32b pm entry */ + jmp startup_32 +SYM_FUNC_END(sl_stub) + +SYM_FUNC_START(sl_txt_ap_entry) + cli + cld + /* + * The %cs and %ds segments are known good after waking the AP. + * First order of business is to find where we are and + * save it in %ebx. + */ + + /* Read physical base of heap into EAX */ + movl (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_HEAP_BASE), %eax + /* Read the size of the BIOS data into ECX (first 8 bytes) */ + movl (%eax), %ecx + /* Skip over BIOS data and size of OS to MLE data section */ + leal 8(%eax, %ecx), %eax + + /* Saved %ebx from the BSP and stash OS-MLE pointer */ + movl (SL_mle_scratch + SL_SCRATCH_AP_EBX)(%eax), %ebx + /* Save OS-MLE base in %edi for call to sl_txt_load_regs */ + movl %eax, %edi + + /* Lock and get our stack index */ + movl $1, %ecx +.Lspin: + xorl %eax, %eax + lock cmpxchgl %ecx, sl_txt_spin_lock(%ebx) + pause + jnz .Lspin + + /* Increment the stack index and use the next value inside lock */ + incl sl_txt_stack_index(%ebx) + movl sl_txt_stack_index(%ebx), %eax + + /* Unlock */ + movl $0, sl_txt_spin_lock(%ebx) + + /* Location of the relocated AP wake block */ + movl sl_txt_ap_wake_block(%ebx), %ecx + + /* Load reloc GDT, set segment regs and lret to __SL32_CS */ + lgdt (sl_ap_gdt_desc - sl_txt_ap_wake_begin)(%ecx) + + movl $(__SL32_DS), %edx + movw %dx, %ds + movw %dx, %es + movw %dx, %fs + movw %dx, %gs + movw %dx, %ss + + /* Load our reloc AP stack */ + movl $(TXT_BOOT_STACK_SIZE), %edx + mull %edx + leal (sl_stacks_end - sl_txt_ap_wake_begin)(%ecx), %esp + subl %eax, %esp + + /* Switch to AP code segment */ + leal .Lsl_ap_cs(%ebx), %eax + pushl $(__SL32_CS) + pushl %eax + lret + +.Lsl_ap_cs: + /* Load the relocated AP IDT */ + lidt (sl_ap_idt_desc - sl_txt_ap_wake_begin)(%ecx) + + /* Fixup MTRRs and misc enable MSR on APs too */ + call sl_txt_load_regs + + /* Enable SMI with GETSEC[SMCTRL] */ + GETSEC $(SMX_X86_GETSEC_SMCTRL) + + /* IRET-to-self can be used to enable NMIs which SENTER disabled */ + leal .Lnmi_enabled_ap(%ebx), %eax + pushfl + pushl $(__SL32_CS) + pushl %eax + iret + +.Lnmi_enabled_ap: + /* Put APs in X2APIC mode like the BSP */ + movl $(MSR_IA32_APICBASE), %ecx + rdmsr + orl $(XAPIC_ENABLE | X2APIC_ENABLE), %eax + wrmsr + + /* + * Basically done, increment the CPU count and jump off to the AP + * wake block to wait. + */ + lock incl sl_txt_cpu_count(%ebx) + + movl sl_txt_ap_wake_block(%ebx), %eax + jmp *%eax +SYM_FUNC_END(sl_txt_ap_entry) + +SYM_FUNC_START(sl_txt_reloc_ap_wake) + /* Save boot params register */ + pushl %esi + + movl sl_txt_ap_wake_block(%ebx), %edi + + /* Fixup AP IDT and GDT descriptor before relocating */ + addl %edi, (sl_ap_idt_desc + 2)(%ebx) + addl %edi, (sl_ap_gdt_desc + 2)(%ebx) + + /* + * Copy the AP wake code and AP GDT/IDT to the protected wake block + * provided by the loader. Destination already in %edi. + */ + movl $(sl_txt_ap_wake_end - sl_txt_ap_wake_begin), %ecx + leal sl_txt_ap_wake_begin(%ebx), %esi + rep movsb + + /* Setup the IDT for the APs to use in the relocation block */ + movl sl_txt_ap_wake_block(%ebx), %ecx + addl $(sl_ap_idt - sl_txt_ap_wake_begin), %ecx + xorl %edx, %edx + + /* Form the default reset vector relocation address */ + movl sl_txt_ap_wake_block(%ebx), %esi + addl $(sl_txt_int_reset - sl_txt_ap_wake_begin), %esi + +1: + cmpw $(NR_VECTORS), %dx + jz .Lap_idt_done + + cmpw $(X86_TRAP_NMI), %dx + jz 2f + + /* Load all other fixed vectors with reset handler */ + movl %esi, %eax + movw %ax, (IDT_VECTOR_LO_BITS)(%ecx) + shrl $16, %eax + movw %ax, (IDT_VECTOR_HI_BITS)(%ecx) + jmp 3f + +2: + /* Load single wake NMI IPI vector at the relocation address */ + movl sl_txt_ap_wake_block(%ebx), %eax + addl $(sl_txt_int_ipi_wake - sl_txt_ap_wake_begin), %eax + movw %ax, (IDT_VECTOR_LO_BITS)(%ecx) + shrl $16, %eax + movw %ax, (IDT_VECTOR_HI_BITS)(%ecx) + +3: + incw %dx + addl $8, %ecx + jmp 1b + +.Lap_idt_done: + popl %esi + ret +SYM_FUNC_END(sl_txt_reloc_ap_wake) + +SYM_FUNC_START(sl_txt_load_regs) + /* Save base pointer register */ + pushl %ebx + + /* + * On Intel, the original variable MTRRs and Misc Enable MSR are + * restored on the BSP at early boot. Each AP will also restore + * its MTRRs and Misc Enable MSR. + */ + pushl %edi + addl $(SL_saved_bsp_mtrrs), %edi + movl (%edi), %ebx + pushl %ebx /* default_mem_type lo */ + addl $4, %edi + movl (%edi), %ebx + pushl %ebx /* default_mem_type hi */ + addl $4, %edi + movl (%edi), %ebx /* mtrr_vcnt lo, don't care about hi part */ + addl $8, %edi /* now at MTRR pair array */ + /* Write the variable MTRRs */ + movl $(MTRRphysBase0), %ecx +1: + cmpl $0, %ebx + jz 2f + + movl (%edi), %eax /* MTRRphysBaseX lo */ + addl $4, %edi + movl (%edi), %edx /* MTRRphysBaseX hi */ + wrmsr + addl $4, %edi + incl %ecx + movl (%edi), %eax /* MTRRphysMaskX lo */ + addl $4, %edi + movl (%edi), %edx /* MTRRphysMaskX hi */ + wrmsr + addl $4, %edi + incl %ecx + + decl %ebx + jmp 1b +2: + /* Write the default MTRR register */ + popl %edx + popl %eax + movl $(MSR_MTRRdefType), %ecx + wrmsr + + /* Return to beginning and write the misc enable msr */ + popl %edi + addl $(SL_saved_misc_enable_msr), %edi + movl (%edi), %eax /* saved_misc_enable_msr lo */ + addl $4, %edi + movl (%edi), %edx /* saved_misc_enable_msr hi */ + movl $(MSR_IA32_MISC_ENABLE), %ecx + wrmsr + + popl %ebx + ret +SYM_FUNC_END(sl_txt_load_regs) + +SYM_FUNC_START(sl_txt_wake_aps) + /* Save boot params register */ + pushl %esi + + /* First setup the MLE join structure and load it into TXT reg */ + leal sl_gdt(%ebx), %eax + leal sl_txt_ap_entry(%ebx), %ecx + leal sl_smx_rlp_mle_join(%ebx), %edx + movl %eax, SL_rlp_gdt_base(%edx) + movl %ecx, SL_rlp_entry_point(%edx) + movl %edx, (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_MLE_JOIN) + + /* Another TXT heap walk to find various values needed to wake APs */ + movl (TXT_PRIV_CONFIG_REGS_BASE + TXT_CR_HEAP_BASE), %eax + /* At BIOS data size, find the number of logical processors */ + movl (SL_num_logical_procs + 8)(%eax), %edx + /* Skip over BIOS data */ + movl (%eax), %ecx + addl %ecx, %eax + /* Skip over OS to MLE */ + movl (%eax), %ecx + addl %ecx, %eax + /* At OS-SNIT size, get capabilities to know how to wake up the APs */ + movl (SL_capabilities + 8)(%eax), %esi + /* Skip over OS to SNIT */ + movl (%eax), %ecx + addl %ecx, %eax + /* At SINIT-MLE size, get the AP wake MONITOR address */ + movl (SL_rlp_wakeup_addr + 8)(%eax), %edi + + /* Determine how to wake up the APs */ + testl $(1 << TXT_SINIT_MLE_CAP_WAKE_MONITOR), %esi + jz .Lwake_getsec + + /* Wake using MWAIT MONITOR */ + movl $1, (%edi) + jmp .Laps_awake + +.Lwake_getsec: + /* Wake using GETSEC(WAKEUP) */ + GETSEC $(SMX_X86_GETSEC_WAKEUP) + +.Laps_awake: + /* + * All of the APs are woken up and rendesvous in the relocated wake + * block starting at sl_txt_ap_wake_begin. Wait for all of them to + * halt. + */ + pause + cmpl sl_txt_cpu_count(%ebx), %edx + jne .Laps_awake + + popl %esi + ret +SYM_FUNC_END(sl_txt_wake_aps) + +/* This is the beginning of the relocated AP wake code block */ + .global sl_txt_ap_wake_begin +sl_txt_ap_wake_begin: + + /* + * Wait for NMI IPI in the relocated AP wake block which was provided + * and protected in the memory map by the prelaunch code. Leave all + * other interrupts masked since we do not expect anything but an NMI. + */ + xorl %edx, %edx + +1: + hlt + testl %edx, %edx + jz 1b + + /* + * This is the long absolute jump to the 32b Secure Launch protected + * mode stub code in the rmpiggy. The jump address will be fixed in + * the SMP boot code when the first AP is brought up. This whole area + * is provided and protected in the memory map by the prelaunch code. + */ + .byte 0xea +sl_ap_jmp_offset: + .long 0x00000000 + .word __SL32_CS + +SYM_FUNC_START(sl_txt_int_ipi_wake) + movl $1, %edx + + /* NMI context, just IRET */ + iret +SYM_FUNC_END(sl_txt_int_ipi_wake) + +SYM_FUNC_START(sl_txt_int_reset) + TXT_RESET $(SL_ERROR_INV_AP_INTERRUPT) +SYM_FUNC_END(sl_txt_int_reset) + + .balign 16 +sl_ap_idt_desc: + .word sl_ap_idt_end - sl_ap_idt - 1 /* Limit */ + .long sl_ap_idt - sl_txt_ap_wake_begin /* Base */ +sl_ap_idt_desc_end: + + .balign 16 +sl_ap_idt: + .rept NR_VECTORS + .word 0x0000 /* Offset 15 to 0 */ + .word __SL32_CS /* Segment selector */ + .word 0x8e00 /* Present, DPL=0, 32b Vector, Interrupt */ + .word 0x0000 /* Offset 31 to 16 */ + .endr +sl_ap_idt_end: + + .balign 16 +sl_ap_gdt_desc: + .word sl_ap_gdt_end - sl_ap_gdt - 1 + .long sl_ap_gdt - sl_txt_ap_wake_begin +sl_ap_gdt_desc_end: + + .balign 16 +sl_ap_gdt: + .quad 0x0000000000000000 /* NULL */ + .quad 0x00cf9a000000ffff /* __SL32_CS */ + .quad 0x00cf92000000ffff /* __SL32_DS */ +sl_ap_gdt_end: + + /* Small stacks for BSP and APs to work with */ + .balign 4 +sl_stacks: + .fill (TXT_MAX_CPUS * TXT_BOOT_STACK_SIZE), 1, 0 +sl_stacks_end: + +/* This is the end of the relocated AP wake code block */ + .global sl_txt_ap_wake_end +sl_txt_ap_wake_end: + + .data + .balign 16 +sl_gdt_desc: + .word sl_gdt_end - sl_gdt - 1 + .long sl_gdt +sl_gdt_desc_end: + + .balign 16 +sl_gdt: + .quad 0x0000000000000000 /* NULL */ + .quad 0x00cf9a000000ffff /* __SL32_CS */ + .quad 0x00cf92000000ffff /* __SL32_DS */ +sl_gdt_end: + + .balign 16 +sl_smx_rlp_mle_join: + .long sl_gdt_end - sl_gdt - 1 /* GDT limit */ + .long 0x00000000 /* GDT base */ + .long __SL32_CS /* Seg Sel - CS (DS, ES, SS = seg_sel+8) */ + .long 0x00000000 /* Entry point physical address */ + +SYM_DATA_START(sl_cpu_type) + .long 0x00000000 +SYM_DATA_END(sl_cpu_type) + +sl_txt_spin_lock: + .long 0x00000000 + +sl_txt_stack_index: + .long 0x00000000 + +sl_txt_cpu_count: + .long 0x00000000 + +sl_txt_ap_wake_block: + .long 0x00000000 diff --git a/arch/x86/kernel/asm-offsets.c b/arch/x86/kernel/asm-offsets.c index 3ca07ad..c7e5ed1 100644 --- a/arch/x86/kernel/asm-offsets.c +++ b/arch/x86/kernel/asm-offsets.c @@ -12,6 +12,7 @@ #include <linux/hardirq.h> #include <linux/suspend.h> #include <linux/kbuild.h> +#include <linux/slaunch.h> #include <asm/processor.h> #include <asm/thread_info.h> #include <asm/sigframe.h> @@ -104,4 +105,19 @@ static void __used common(void) OFFSET(TSS_sp0, tss_struct, x86_tss.sp0); OFFSET(TSS_sp1, tss_struct, x86_tss.sp1); OFFSET(TSS_sp2, tss_struct, x86_tss.sp2); + +#ifdef CONFIG_SECURE_LAUNCH + BLANK(); + OFFSET(SL_boot_params_addr, txt_os_mle_data, boot_params_addr); + OFFSET(SL_saved_misc_enable_msr, txt_os_mle_data, saved_misc_enable_msr); + OFFSET(SL_saved_bsp_mtrrs, txt_os_mle_data, saved_bsp_mtrrs); + OFFSET(SL_ap_wake_block, txt_os_mle_data, ap_wake_block); + OFFSET(SL_ap_wake_block_size, txt_os_mle_data, ap_wake_block_size); + OFFSET(SL_mle_scratch, txt_os_mle_data, mle_scratch); + OFFSET(SL_num_logical_procs, txt_bios_data, num_logical_procs); + OFFSET(SL_capabilities, txt_os_sinit_data, capabilities); + OFFSET(SL_rlp_wakeup_addr, txt_sinit_mle_data, rlp_wakeup_addr); + OFFSET(SL_rlp_gdt_base, smx_rlp_mle_join, rlp_gdt_base); + OFFSET(SL_rlp_entry_point, smx_rlp_mle_join, rlp_entry_point); +#endif }
The Secure Launch (SL) stub provides the entry point for Intel TXT (and later AMD SKINIT) to vector to during the late launch. The symbol sl_stub_entry is that entry point and its offset into the kernel is conveyed to the launching code using the MLE (Measured Launch Environment) header in the structure named mle_header. The offset of the MLE header is set in the kernel_info. The routine sl_stub contains the very early late launch setup code responsible for setting up the basic environment to allow the normal kernel startup_32 code to proceed. It is also responsible for properly waking and handling the APs on Intel platforms. The routine sl_main which runs after entering 64b mode is responsible for measuring configuration and module information before it is used like the boot params, the kernel command line, the TXT heap, an external initramfs, etc. Signed-off-by: Ross Philipson <ross.philipson@oracle.com> --- Documentation/x86/boot.rst | 9 + arch/x86/boot/compressed/Makefile | 1 + arch/x86/boot/compressed/head_64.S | 34 ++ arch/x86/boot/compressed/kernel_info.S | 7 + arch/x86/boot/compressed/sl_main.c | 390 +++++++++++++++++++++ arch/x86/boot/compressed/sl_stub.S | 606 +++++++++++++++++++++++++++++++++ arch/x86/kernel/asm-offsets.c | 16 + 7 files changed, 1063 insertions(+) create mode 100644 arch/x86/boot/compressed/sl_main.c create mode 100644 arch/x86/boot/compressed/sl_stub.S