| Message ID | 20240306085622.87248-3-cuiyunhui@bytedance.com (mailing list archive) |
|---|---|
| State | Changes Requested |
| Headers | show |
| Series | [1/3] Revert "riscv/efistub: Ensure GP-relative addressing is not used" | expand |
| Context | Check | Description |
|---|---|---|
| conchuod/vmtest-fixes-PR | fail | merge-conflict |
On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: > > Compared with gcc version 12, gcc version 13 uses the gp > register for compilation optimization, but the efistub module > does not initialize gp. > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> This needs a sign-off, and your signoff needs to come after. > --- > arch/riscv/kernel/efi-header.S | 11 ++++++++++- > 1 file changed, 10 insertions(+), 1 deletion(-) > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S > index 515b2dfbca75..fa17c08c092a 100644 > --- a/arch/riscv/kernel/efi-header.S > +++ b/arch/riscv/kernel/efi-header.S > @@ -40,7 +40,7 @@ optional_header: > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData > #endif > .long 0 // SizeOfUninitializedData > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint > + .long _efistub_entry - _start // AddressOfEntryPoint > .long efi_header_end - _start // BaseOfCode > #ifdef CONFIG_32BIT > .long __pecoff_text_end - _start // BaseOfData > @@ -121,4 +121,13 @@ section_table: > > .balign 0x1000 > efi_header_end: > + > + .global _efistub_entry > +_efistub_entry: This should go into .text or .init.text, not the header. > + /* Reload the global pointer */ > + load_global_pointer > + What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI stub Makefile removes the SCS CFLAGS, so the stub will be built without shadow call stack support, which I guess means that it might use GP as a global pointer as usual? > + call __efistub_efi_pe_entry > + ret > + You are returning to the firmware here, but after modifying the GP register. Shouldn't you restore it to its old value?
Hi Ard, On Wed, Mar 6, 2024 at 5:36 PM Ard Biesheuvel <ardb@kernel.org> wrote: > > On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: > > > > Compared with gcc version 12, gcc version 13 uses the gp > > register for compilation optimization, but the efistub module > > does not initialize gp. > > > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> > > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> > > This needs a sign-off, and your signoff needs to come after. > > > --- > > arch/riscv/kernel/efi-header.S | 11 ++++++++++- > > 1 file changed, 10 insertions(+), 1 deletion(-) > > > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S > > index 515b2dfbca75..fa17c08c092a 100644 > > --- a/arch/riscv/kernel/efi-header.S > > +++ b/arch/riscv/kernel/efi-header.S > > @@ -40,7 +40,7 @@ optional_header: > > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData > > #endif > > .long 0 // SizeOfUninitializedData > > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint > > + .long _efistub_entry - _start // AddressOfEntryPoint > > .long efi_header_end - _start // BaseOfCode > > #ifdef CONFIG_32BIT > > .long __pecoff_text_end - _start // BaseOfData > > @@ -121,4 +121,13 @@ section_table: > > > > .balign 0x1000 > > efi_header_end: > > + > > + .global _efistub_entry > > +_efistub_entry: > > This should go into .text or .init.text, not the header. > > > + /* Reload the global pointer */ > > + load_global_pointer > > + > > What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI > stub Makefile removes the SCS CFLAGS, so the stub will be built > without shadow call stack support, which I guess means that it might > use GP as a global pointer as usual? > > > + call __efistub_efi_pe_entry > > + ret > > + > > You are returning to the firmware here, but after modifying the GP > register. Shouldn't you restore it to its old value? There is no need to restore the value of the gp register. Where gp is needed, the gp register must first be initialized. And here is the entry. Regarding your first two comments above, I plan to make the following changes in v2, efi_header_end: + + __INIT + .global _efistub_entry +_efistub_entry: + /* Reload the global pointer */ +.option push +.option norelax + la gp, __global_pointer$ +.option pop + + call __efistub_efi_pe_entry + ret + __HEAD + .endm what do you think? Thanks, Yunhui
On Wed, 6 Mar 2024 at 13:34, yunhui cui <cuiyunhui@bytedance.com> wrote: > > Hi Ard, > > On Wed, Mar 6, 2024 at 5:36 PM Ard Biesheuvel <ardb@kernel.org> wrote: > > > > On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: > > > > > > Compared with gcc version 12, gcc version 13 uses the gp > > > register for compilation optimization, but the efistub module > > > does not initialize gp. > > > > > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> > > > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> > > > > This needs a sign-off, and your signoff needs to come after. > > > > > --- > > > arch/riscv/kernel/efi-header.S | 11 ++++++++++- > > > 1 file changed, 10 insertions(+), 1 deletion(-) > > > > > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S > > > index 515b2dfbca75..fa17c08c092a 100644 > > > --- a/arch/riscv/kernel/efi-header.S > > > +++ b/arch/riscv/kernel/efi-header.S > > > @@ -40,7 +40,7 @@ optional_header: > > > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData > > > #endif > > > .long 0 // SizeOfUninitializedData > > > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint > > > + .long _efistub_entry - _start // AddressOfEntryPoint > > > .long efi_header_end - _start // BaseOfCode > > > #ifdef CONFIG_32BIT > > > .long __pecoff_text_end - _start // BaseOfData > > > @@ -121,4 +121,13 @@ section_table: > > > > > > .balign 0x1000 > > > efi_header_end: > > > + > > > + .global _efistub_entry > > > +_efistub_entry: > > > > This should go into .text or .init.text, not the header. > > > > > + /* Reload the global pointer */ > > > + load_global_pointer > > > + > > > > What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI > > stub Makefile removes the SCS CFLAGS, so the stub will be built > > without shadow call stack support, which I guess means that it might > > use GP as a global pointer as usual? > > > > > + call __efistub_efi_pe_entry > > > + ret > > > + > > > > You are returning to the firmware here, but after modifying the GP > > register. Shouldn't you restore it to its old value? > There is no need to restore the value of the gp register. Where gp is > needed, the gp register must first be initialized. And here is the > entry. > But how should the firmware know that GP was corrupted after calling the kernel's EFI entrypoint? The EFI stub can return to the firmware if it encounters any errors while still running in the EFI boot services. > Regarding your first two comments above, I plan to make the following > changes in v2, > efi_header_end: > + > + __INIT > + .global _efistub_entry > +_efistub_entry: > + /* Reload the global pointer */ > +.option push > +.option norelax > + la gp, __global_pointer$ > +.option pop > + > + call __efistub_efi_pe_entry > + ret > + __HEAD > + > .endm > > what do you think? > This looks ok to me, but I would still like to understand why it is ok to return to the firmware with a modified GP value.
On Wed, 6 Mar 2024 at 14:02, Ard Biesheuvel <ardb@kernel.org> wrote: > > On Wed, 6 Mar 2024 at 13:34, yunhui cui <cuiyunhui@bytedance.com> wrote: > > > > Hi Ard, > > > > On Wed, Mar 6, 2024 at 5:36 PM Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > > On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: > > > > > > > > Compared with gcc version 12, gcc version 13 uses the gp > > > > register for compilation optimization, but the efistub module > > > > does not initialize gp. > > > > > > > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> > > > > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> > > > > > > This needs a sign-off, and your signoff needs to come after. > > > > > > > --- > > > > arch/riscv/kernel/efi-header.S | 11 ++++++++++- > > > > 1 file changed, 10 insertions(+), 1 deletion(-) > > > > > > > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S > > > > index 515b2dfbca75..fa17c08c092a 100644 > > > > --- a/arch/riscv/kernel/efi-header.S > > > > +++ b/arch/riscv/kernel/efi-header.S > > > > @@ -40,7 +40,7 @@ optional_header: > > > > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData > > > > #endif > > > > .long 0 // SizeOfUninitializedData > > > > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint > > > > + .long _efistub_entry - _start // AddressOfEntryPoint > > > > .long efi_header_end - _start // BaseOfCode > > > > #ifdef CONFIG_32BIT > > > > .long __pecoff_text_end - _start // BaseOfData > > > > @@ -121,4 +121,13 @@ section_table: > > > > > > > > .balign 0x1000 > > > > efi_header_end: > > > > + > > > > + .global _efistub_entry > > > > +_efistub_entry: > > > > > > This should go into .text or .init.text, not the header. > > > > > > > + /* Reload the global pointer */ > > > > + load_global_pointer > > > > + > > > > > > What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI > > > stub Makefile removes the SCS CFLAGS, so the stub will be built > > > without shadow call stack support, which I guess means that it might > > > use GP as a global pointer as usual? > > > > > > > + call __efistub_efi_pe_entry > > > > + ret > > > > + > > > > > > You are returning to the firmware here, but after modifying the GP > > > register. Shouldn't you restore it to its old value? > > There is no need to restore the value of the gp register. Where gp is > > needed, the gp register must first be initialized. And here is the > > entry. > > > > But how should the firmware know that GP was corrupted after calling > the kernel's EFI entrypoint? The EFI stub can return to the firmware > if it encounters any errors while still running in the EFI boot > services. > Actually, I wonder if GP can be modified at all before ExitBootServices(). The EFI timer interrupt is still live at this point, and so the firmware is being called behind your back, and might rely on GP retaining its original value.
Hi Ard, On Wed, Mar 6, 2024 at 9:09 PM Ard Biesheuvel <ardb@kernel.org> wrote: > > On Wed, 6 Mar 2024 at 14:02, Ard Biesheuvel <ardb@kernel.org> wrote: > > > > On Wed, 6 Mar 2024 at 13:34, yunhui cui <cuiyunhui@bytedance.com> wrote: > > > > > > Hi Ard, > > > > > > On Wed, Mar 6, 2024 at 5:36 PM Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > > > > On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: > > > > > > > > > > Compared with gcc version 12, gcc version 13 uses the gp > > > > > register for compilation optimization, but the efistub module > > > > > does not initialize gp. > > > > > > > > > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> > > > > > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> > > > > > > > > This needs a sign-off, and your signoff needs to come after. > > > > > > > > > --- > > > > > arch/riscv/kernel/efi-header.S | 11 ++++++++++- > > > > > 1 file changed, 10 insertions(+), 1 deletion(-) > > > > > > > > > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S > > > > > index 515b2dfbca75..fa17c08c092a 100644 > > > > > --- a/arch/riscv/kernel/efi-header.S > > > > > +++ b/arch/riscv/kernel/efi-header.S > > > > > @@ -40,7 +40,7 @@ optional_header: > > > > > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData > > > > > #endif > > > > > .long 0 // SizeOfUninitializedData > > > > > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint > > > > > + .long _efistub_entry - _start // AddressOfEntryPoint > > > > > .long efi_header_end - _start // BaseOfCode > > > > > #ifdef CONFIG_32BIT > > > > > .long __pecoff_text_end - _start // BaseOfData > > > > > @@ -121,4 +121,13 @@ section_table: > > > > > > > > > > .balign 0x1000 > > > > > efi_header_end: > > > > > + > > > > > + .global _efistub_entry > > > > > +_efistub_entry: > > > > > > > > This should go into .text or .init.text, not the header. > > > > > > > > > + /* Reload the global pointer */ > > > > > + load_global_pointer > > > > > + > > > > > > > > What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI > > > > stub Makefile removes the SCS CFLAGS, so the stub will be built > > > > without shadow call stack support, which I guess means that it might > > > > use GP as a global pointer as usual? > > > > > > > > > + call __efistub_efi_pe_entry > > > > > + ret > > > > > + > > > > > > > > You are returning to the firmware here, but after modifying the GP > > > > register. Shouldn't you restore it to its old value? > > > There is no need to restore the value of the gp register. Where gp is > > > needed, the gp register must first be initialized. And here is the > > > entry. > > > > > > > But how should the firmware know that GP was corrupted after calling > > the kernel's EFI entrypoint? The EFI stub can return to the firmware > > if it encounters any errors while still running in the EFI boot > > services. > > > > Actually, I wonder if GP can be modified at all before > ExitBootServices(). The EFI timer interrupt is still live at this > point, and so the firmware is being called behind your back, and might > rely on GP retaining its original value. OK, in v2 I will restore the value of gp as follows: efi_header_end: + + __INIT + .global_efistub_entry +_efistub_entry: + /* Reload the global pointer */ +.option push +.option norelax + addi sp,sp,-8 + sd gp,0(sp) + la gp, __global_pointer$ +.option pop + call __efistub_efi_pe_entry +ld gp,0(sp) +addi sp,sp,8 +ret + __HEAD + .endm what do you think? Thanks, Yunhui
On Wed, 06 Mar 2024 05:09:07 PST (-0800), Ard Biesheuvel wrote: > On Wed, 6 Mar 2024 at 14:02, Ard Biesheuvel <ardb@kernel.org> wrote: >> >> On Wed, 6 Mar 2024 at 13:34, yunhui cui <cuiyunhui@bytedance.com> wrote: >> > >> > Hi Ard, >> > >> > On Wed, Mar 6, 2024 at 5:36 PM Ard Biesheuvel <ardb@kernel.org> wrote: >> > > >> > > On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: >> > > > >> > > > Compared with gcc version 12, gcc version 13 uses the gp >> > > > register for compilation optimization, but the efistub module >> > > > does not initialize gp. >> > > > >> > > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> >> > > > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> >> > > >> > > This needs a sign-off, and your signoff needs to come after. >> > > >> > > > --- >> > > > arch/riscv/kernel/efi-header.S | 11 ++++++++++- >> > > > 1 file changed, 10 insertions(+), 1 deletion(-) >> > > > >> > > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S >> > > > index 515b2dfbca75..fa17c08c092a 100644 >> > > > --- a/arch/riscv/kernel/efi-header.S >> > > > +++ b/arch/riscv/kernel/efi-header.S >> > > > @@ -40,7 +40,7 @@ optional_header: >> > > > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData >> > > > #endif >> > > > .long 0 // SizeOfUninitializedData >> > > > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint >> > > > + .long _efistub_entry - _start // AddressOfEntryPoint >> > > > .long efi_header_end - _start // BaseOfCode >> > > > #ifdef CONFIG_32BIT >> > > > .long __pecoff_text_end - _start // BaseOfData >> > > > @@ -121,4 +121,13 @@ section_table: >> > > > >> > > > .balign 0x1000 >> > > > efi_header_end: >> > > > + >> > > > + .global _efistub_entry >> > > > +_efistub_entry: >> > > >> > > This should go into .text or .init.text, not the header. >> > > >> > > > + /* Reload the global pointer */ >> > > > + load_global_pointer >> > > > + >> > > >> > > What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI >> > > stub Makefile removes the SCS CFLAGS, so the stub will be built >> > > without shadow call stack support, which I guess means that it might >> > > use GP as a global pointer as usual? >> > > >> > > > + call __efistub_efi_pe_entry >> > > > + ret >> > > > + >> > > >> > > You are returning to the firmware here, but after modifying the GP >> > > register. Shouldn't you restore it to its old value? >> > There is no need to restore the value of the gp register. Where gp is >> > needed, the gp register must first be initialized. And here is the >> > entry. >> > >> >> But how should the firmware know that GP was corrupted after calling >> the kernel's EFI entrypoint? The EFI stub can return to the firmware >> if it encounters any errors while still running in the EFI boot >> services. >> > > Actually, I wonder if GP can be modified at all before > ExitBootServices(). The EFI timer interrupt is still live at this > point, and so the firmware is being called behind your back, and might > rely on GP retaining its original value. [A few of us are talking on IRC as I'm writing this...] The UEFI spec says "UEFI firmware must neither trust the values of tp and gp nor make an assumption of owning the write access to these register in any circumstances". It's kind of vague what "UEFI firmware" means here, but I think it's reasonable to assume that the kernel (and thus the EFI stub) is not included there. So under that interpretation, the kernel (including the EFI stub) would be allowed to overwrite GP with whatever it wants. [We're still talking on IRC, though]
On Wed, 6 Mar 2024 at 16:21, Palmer Dabbelt <palmer@dabbelt.com> wrote: > > On Wed, 06 Mar 2024 05:09:07 PST (-0800), Ard Biesheuvel wrote: > > On Wed, 6 Mar 2024 at 14:02, Ard Biesheuvel <ardb@kernel.org> wrote: > >> > >> On Wed, 6 Mar 2024 at 13:34, yunhui cui <cuiyunhui@bytedance.com> wrote: > >> > > >> > Hi Ard, > >> > > >> > On Wed, Mar 6, 2024 at 5:36 PM Ard Biesheuvel <ardb@kernel.org> wrote: > >> > > > >> > > On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: > >> > > > > >> > > > Compared with gcc version 12, gcc version 13 uses the gp > >> > > > register for compilation optimization, but the efistub module > >> > > > does not initialize gp. > >> > > > > >> > > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> > >> > > > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> > >> > > > >> > > This needs a sign-off, and your signoff needs to come after. > >> > > > >> > > > --- > >> > > > arch/riscv/kernel/efi-header.S | 11 ++++++++++- > >> > > > 1 file changed, 10 insertions(+), 1 deletion(-) > >> > > > > >> > > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S > >> > > > index 515b2dfbca75..fa17c08c092a 100644 > >> > > > --- a/arch/riscv/kernel/efi-header.S > >> > > > +++ b/arch/riscv/kernel/efi-header.S > >> > > > @@ -40,7 +40,7 @@ optional_header: > >> > > > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData > >> > > > #endif > >> > > > .long 0 // SizeOfUninitializedData > >> > > > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint > >> > > > + .long _efistub_entry - _start // AddressOfEntryPoint > >> > > > .long efi_header_end - _start // BaseOfCode > >> > > > #ifdef CONFIG_32BIT > >> > > > .long __pecoff_text_end - _start // BaseOfData > >> > > > @@ -121,4 +121,13 @@ section_table: > >> > > > > >> > > > .balign 0x1000 > >> > > > efi_header_end: > >> > > > + > >> > > > + .global _efistub_entry > >> > > > +_efistub_entry: > >> > > > >> > > This should go into .text or .init.text, not the header. > >> > > > >> > > > + /* Reload the global pointer */ > >> > > > + load_global_pointer > >> > > > + > >> > > > >> > > What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI > >> > > stub Makefile removes the SCS CFLAGS, so the stub will be built > >> > > without shadow call stack support, which I guess means that it might > >> > > use GP as a global pointer as usual? > >> > > > >> > > > + call __efistub_efi_pe_entry > >> > > > + ret > >> > > > + > >> > > > >> > > You are returning to the firmware here, but after modifying the GP > >> > > register. Shouldn't you restore it to its old value? > >> > There is no need to restore the value of the gp register. Where gp is > >> > needed, the gp register must first be initialized. And here is the > >> > entry. > >> > > >> > >> But how should the firmware know that GP was corrupted after calling > >> the kernel's EFI entrypoint? The EFI stub can return to the firmware > >> if it encounters any errors while still running in the EFI boot > >> services. > >> > > > > Actually, I wonder if GP can be modified at all before > > ExitBootServices(). The EFI timer interrupt is still live at this > > point, and so the firmware is being called behind your back, and might > > rely on GP retaining its original value. > > [A few of us are talking on IRC as I'm writing this...] > > The UEFI spec says "UEFI firmware must neither trust the > values of tp and gp nor make an assumption of owning the write access to > these register in any circumstances". It's kind of vague what "UEFI > firmware" means here, but I think it's reasonable to assume that the > kernel (and thus the EFI stub) is not included there. > > So under that interpretation, the kernel (including the EFI stub) would > be allowed to overwrite GP with whatever it wants. > OK, so even if the UEFI spec seems to suggest that using GP in EFI applications such as the Linux EFI stub should be safe, I'd still like to understand why this change is necessary. The patches you are reverting are supposed to ensure that a) the compiler does not generate references that can be relaxed to GP based ones, and b) no R_RISCV_RELAX relocations are present in any of the code that runs in the context of the EFI firmware. Are you still seeing GP based symbol references? Is there C code that gets pulled into the EFI stub that uses GP based relocations perhaps? (see list below). If any of those are implemented in C, they should not be used by the EFI stub directly unless they are guaranteed to be uninstrumented and callable at arbitrary offsets other than the one they were linked to run at. __efistub_memcmp = memcmp; __efistub_memchr = memchr; __efistub_memcpy = memcpy; __efistub_memmove = memmove; __efistub_memset = memset; __efistub_strlen = strlen; __efistub_strnlen = strnlen; __efistub_strcmp = strcmp; __efistub_strncmp = strncmp; __efistub_strrchr = strrchr; __efistub___memcpy = memcpy; __efistub___memmove = memmove; __efistub___memset = memset; __efistub__start = _start; __efistub__start_kernel = _start_kernel; (from arch/riscv/kernel/image-vars.h)
On Wed, 6 Mar 2024 at 16:44, Ard Biesheuvel <ardb@kernel.org> wrote: > > On Wed, 6 Mar 2024 at 16:21, Palmer Dabbelt <palmer@dabbelt.com> wrote: > > > > On Wed, 06 Mar 2024 05:09:07 PST (-0800), Ard Biesheuvel wrote: > > > On Wed, 6 Mar 2024 at 14:02, Ard Biesheuvel <ardb@kernel.org> wrote: > > >> > > >> On Wed, 6 Mar 2024 at 13:34, yunhui cui <cuiyunhui@bytedance.com> wrote: > > >> > > > >> > Hi Ard, > > >> > > > >> > On Wed, Mar 6, 2024 at 5:36 PM Ard Biesheuvel <ardb@kernel.org> wrote: > > >> > > > > >> > > On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: > > >> > > > > > >> > > > Compared with gcc version 12, gcc version 13 uses the gp > > >> > > > register for compilation optimization, but the efistub module > > >> > > > does not initialize gp. > > >> > > > > > >> > > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> > > >> > > > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> > > >> > > > > >> > > This needs a sign-off, and your signoff needs to come after. > > >> > > > > >> > > > --- > > >> > > > arch/riscv/kernel/efi-header.S | 11 ++++++++++- > > >> > > > 1 file changed, 10 insertions(+), 1 deletion(-) > > >> > > > > > >> > > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S > > >> > > > index 515b2dfbca75..fa17c08c092a 100644 > > >> > > > --- a/arch/riscv/kernel/efi-header.S > > >> > > > +++ b/arch/riscv/kernel/efi-header.S > > >> > > > @@ -40,7 +40,7 @@ optional_header: > > >> > > > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData > > >> > > > #endif > > >> > > > .long 0 // SizeOfUninitializedData > > >> > > > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint > > >> > > > + .long _efistub_entry - _start // AddressOfEntryPoint > > >> > > > .long efi_header_end - _start // BaseOfCode > > >> > > > #ifdef CONFIG_32BIT > > >> > > > .long __pecoff_text_end - _start // BaseOfData > > >> > > > @@ -121,4 +121,13 @@ section_table: > > >> > > > > > >> > > > .balign 0x1000 > > >> > > > efi_header_end: > > >> > > > + > > >> > > > + .global _efistub_entry > > >> > > > +_efistub_entry: > > >> > > > > >> > > This should go into .text or .init.text, not the header. > > >> > > > > >> > > > + /* Reload the global pointer */ > > >> > > > + load_global_pointer > > >> > > > + > > >> > > > > >> > > What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI > > >> > > stub Makefile removes the SCS CFLAGS, so the stub will be built > > >> > > without shadow call stack support, which I guess means that it might > > >> > > use GP as a global pointer as usual? > > >> > > > > >> > > > + call __efistub_efi_pe_entry > > >> > > > + ret > > >> > > > + > > >> > > > > >> > > You are returning to the firmware here, but after modifying the GP > > >> > > register. Shouldn't you restore it to its old value? > > >> > There is no need to restore the value of the gp register. Where gp is > > >> > needed, the gp register must first be initialized. And here is the > > >> > entry. > > >> > > > >> > > >> But how should the firmware know that GP was corrupted after calling > > >> the kernel's EFI entrypoint? The EFI stub can return to the firmware > > >> if it encounters any errors while still running in the EFI boot > > >> services. > > >> > > > > > > Actually, I wonder if GP can be modified at all before > > > ExitBootServices(). The EFI timer interrupt is still live at this > > > point, and so the firmware is being called behind your back, and might > > > rely on GP retaining its original value. > > > > [A few of us are talking on IRC as I'm writing this...] > > > > The UEFI spec says "UEFI firmware must neither trust the > > values of tp and gp nor make an assumption of owning the write access to > > these register in any circumstances". It's kind of vague what "UEFI > > firmware" means here, but I think it's reasonable to assume that the > > kernel (and thus the EFI stub) is not included there. > > > > So under that interpretation, the kernel (including the EFI stub) would > > be allowed to overwrite GP with whatever it wants. > > > > OK, so even if the UEFI spec seems to suggest that using GP in EFI > applications such as the Linux EFI stub should be safe, I'd still like > to understand why this change is necessary. The patches you are > reverting are supposed to ensure that a) the compiler does not > generate references that can be relaxed to GP based ones, and b) no > R_RISCV_RELAX relocations are present in any of the code that runs in > the context of the EFI firmware. > > Are you still seeing GP based symbol references? Is there C code that > gets pulled into the EFI stub that uses GP based relocations perhaps? > (see list below). If any of those are implemented in C, they should > not be used by the EFI stub directly unless they are guaranteed to be > uninstrumented and callable at arbitrary offsets other than the one > they were linked to run at. > > > __efistub_memcmp = memcmp; > __efistub_memchr = memchr; > __efistub_memcpy = memcpy; > __efistub_memmove = memmove; > __efistub_memset = memset; > __efistub_strlen = strlen; > __efistub_strnlen = strnlen; > __efistub_strcmp = strcmp; > __efistub_strncmp = strncmp; > __efistub_strrchr = strrchr; > __efistub___memcpy = memcpy; > __efistub___memmove = memmove; > __efistub___memset = memset; > __efistub__start = _start; > __efistub__start_kernel = _start_kernel; > > (from arch/riscv/kernel/image-vars.h) Uhm never mind - these are all gone now, I was looking at a v6.1 kernel source tree. So that means that, as far as I can tell, the only kernel C code that executes in the context of the EFI firmware is built with -mno-relax and is checked for the absence of R_RISCV_RELAX relocations. So I fail to see why these changes are needed. Yunhui, could you please explain the reason for this series?
On Wed, 6 Mar 2024 at 17:15, Ard Biesheuvel <ardb@kernel.org> wrote: > > On Wed, 6 Mar 2024 at 16:44, Ard Biesheuvel <ardb@kernel.org> wrote: > > > > On Wed, 6 Mar 2024 at 16:21, Palmer Dabbelt <palmer@dabbelt.com> wrote: > > > ... > > > > > > The UEFI spec says "UEFI firmware must neither trust the > > > values of tp and gp nor make an assumption of owning the write access to > > > these register in any circumstances". It's kind of vague what "UEFI > > > firmware" means here, but I think it's reasonable to assume that the > > > kernel (and thus the EFI stub) is not included there. > > > > > > So under that interpretation, the kernel (including the EFI stub) would > > > be allowed to overwrite GP with whatever it wants. > > > ... After some more consideration, I concluded that using GP in code that executes in the context of EFI is never safe. Taking the typical Linux/EFI boot sequence as an example, where GRUB is loaded by the system firmware, and the Linux EFI stub is loaded by GRUB, it is easy to spot the problem: GRUB exposes the initial ramdisk via the LoadFile2 protocol, which is essentially a callback interface. If we assume that EFI apps can use GP as they like, we cannot safely call this callback interface without restoring GP to its original value, since we have no idea whether GRUB (or some other loader) is relying on its value. And in addition to synchronous callbacks, there may be EFI event callbacks registered by GRUB that are signaled by the system firmware asynchronously, which means GRUB code could be called behind our backs. Without any guidance in the UEFI spec on how GP needs to be managed across such boundaries, we should assume that GP needs to be restored to the old value in each of those cases, but this is impossible for async event handlers. We might conclude then that using GP is only safe for EFI apps if they don't install protocols or register for EFI events, but that would still imply that we should restore GP to its old value upon exit. So without a strong argument why GP relaxations need to be supported in the EFI stub, I don't think we should consider these changes. If there are any issues in the current code that result in inadvertent GP relaxations, we should address those instead.
Hi Ard, On Thu, Mar 7, 2024 at 12:15 AM Ard Biesheuvel <ardb@kernel.org> wrote: > > On Wed, 6 Mar 2024 at 16:44, Ard Biesheuvel <ardb@kernel.org> wrote: > > > > On Wed, 6 Mar 2024 at 16:21, Palmer Dabbelt <palmer@dabbelt.com> wrote: > > > > > > On Wed, 06 Mar 2024 05:09:07 PST (-0800), Ard Biesheuvel wrote: > > > > On Wed, 6 Mar 2024 at 14:02, Ard Biesheuvel <ardb@kernel.org> wrote: > > > >> > > > >> On Wed, 6 Mar 2024 at 13:34, yunhui cui <cuiyunhui@bytedance.com> wrote: > > > >> > > > > >> > Hi Ard, > > > >> > > > > >> > On Wed, Mar 6, 2024 at 5:36 PM Ard Biesheuvel <ardb@kernel.org> wrote: > > > >> > > > > > >> > > On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: > > > >> > > > > > > >> > > > Compared with gcc version 12, gcc version 13 uses the gp > > > >> > > > register for compilation optimization, but the efistub module > > > >> > > > does not initialize gp. > > > >> > > > > > > >> > > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> > > > >> > > > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> > > > >> > > > > > >> > > This needs a sign-off, and your signoff needs to come after. > > > >> > > > > > >> > > > --- > > > >> > > > arch/riscv/kernel/efi-header.S | 11 ++++++++++- > > > >> > > > 1 file changed, 10 insertions(+), 1 deletion(-) > > > >> > > > > > > >> > > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S > > > >> > > > index 515b2dfbca75..fa17c08c092a 100644 > > > >> > > > --- a/arch/riscv/kernel/efi-header.S > > > >> > > > +++ b/arch/riscv/kernel/efi-header.S > > > >> > > > @@ -40,7 +40,7 @@ optional_header: > > > >> > > > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData > > > >> > > > #endif > > > >> > > > .long 0 // SizeOfUninitializedData > > > >> > > > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint > > > >> > > > + .long _efistub_entry - _start // AddressOfEntryPoint > > > >> > > > .long efi_header_end - _start // BaseOfCode > > > >> > > > #ifdef CONFIG_32BIT > > > >> > > > .long __pecoff_text_end - _start // BaseOfData > > > >> > > > @@ -121,4 +121,13 @@ section_table: > > > >> > > > > > > >> > > > .balign 0x1000 > > > >> > > > efi_header_end: > > > >> > > > + > > > >> > > > + .global _efistub_entry > > > >> > > > +_efistub_entry: > > > >> > > > > > >> > > This should go into .text or .init.text, not the header. > > > >> > > > > > >> > > > + /* Reload the global pointer */ > > > >> > > > + load_global_pointer > > > >> > > > + > > > >> > > > > > >> > > What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI > > > >> > > stub Makefile removes the SCS CFLAGS, so the stub will be built > > > >> > > without shadow call stack support, which I guess means that it might > > > >> > > use GP as a global pointer as usual? > > > >> > > > > > >> > > > + call __efistub_efi_pe_entry > > > >> > > > + ret > > > >> > > > + > > > >> > > > > > >> > > You are returning to the firmware here, but after modifying the GP > > > >> > > register. Shouldn't you restore it to its old value? > > > >> > There is no need to restore the value of the gp register. Where gp is > > > >> > needed, the gp register must first be initialized. And here is the > > > >> > entry. > > > >> > > > > >> > > > >> But how should the firmware know that GP was corrupted after calling > > > >> the kernel's EFI entrypoint? The EFI stub can return to the firmware > > > >> if it encounters any errors while still running in the EFI boot > > > >> services. > > > >> > > > > > > > > Actually, I wonder if GP can be modified at all before > > > > ExitBootServices(). The EFI timer interrupt is still live at this > > > > point, and so the firmware is being called behind your back, and might > > > > rely on GP retaining its original value. > > > > > > [A few of us are talking on IRC as I'm writing this...] > > > > > > The UEFI spec says "UEFI firmware must neither trust the > > > values of tp and gp nor make an assumption of owning the write access to > > > these register in any circumstances". It's kind of vague what "UEFI > > > firmware" means here, but I think it's reasonable to assume that the > > > kernel (and thus the EFI stub) is not included there. > > > > > > So under that interpretation, the kernel (including the EFI stub) would > > > be allowed to overwrite GP with whatever it wants. > > > > > > > OK, so even if the UEFI spec seems to suggest that using GP in EFI > > applications such as the Linux EFI stub should be safe, I'd still like > > to understand why this change is necessary. The patches you are > > reverting are supposed to ensure that a) the compiler does not > > generate references that can be relaxed to GP based ones, and b) no > > R_RISCV_RELAX relocations are present in any of the code that runs in > > the context of the EFI firmware. > > > > Are you still seeing GP based symbol references? Is there C code that > > gets pulled into the EFI stub that uses GP based relocations perhaps? > > (see list below). If any of those are implemented in C, they should > > not be used by the EFI stub directly unless they are guaranteed to be > > uninstrumented and callable at arbitrary offsets other than the one > > they were linked to run at. > > > > > > __efistub_memcmp = memcmp; > > __efistub_memchr = memchr; > > __efistub_memcpy = memcpy; > > __efistub_memmove = memmove; > > __efistub_memset = memset; > > __efistub_strlen = strlen; > > __efistub_strnlen = strnlen; > > __efistub_strcmp = strcmp; > > __efistub_strncmp = strncmp; > > __efistub_strrchr = strrchr; > > __efistub___memcpy = memcpy; > > __efistub___memmove = memmove; > > __efistub___memset = memset; > > __efistub__start = _start; > > __efistub__start_kernel = _start_kernel; > > > > (from arch/riscv/kernel/image-vars.h) > > Uhm never mind - these are all gone now, I was looking at a v6.1 > kernel source tree. > > So that means that, as far as I can tell, the only kernel C code that > executes in the context of the EFI firmware is built with -mno-relax > and is checked for the absence of R_RISCV_RELAX relocations. So I fail > to see why these changes are needed. > > Yunhui, could you please explain the reason for this series? From the logic of binutils, if "__global_pointer$" exists, it is possible to use GP for optimization. For RISC-V, "__global_pointer$" was introduced in commit "fbe934d69eb7e". Therefore, for the system as a whole, we should keep using GP uniformly. The root cause of this problem is that GP is not loaded, rather than "On RISC-V, we also avoid GP based relocations..." as commit "d2baf8cc82c17" said. We need to address problems head-on, rather than avoid them. Thanks, Yunhui
On Thu, 7 Mar 2024 at 04:19, yunhui cui <cuiyunhui@bytedance.com> wrote: > > Hi Ard, > > On Thu, Mar 7, 2024 at 12:15 AM Ard Biesheuvel <ardb@kernel.org> wrote: > > > > On Wed, 6 Mar 2024 at 16:44, Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > > On Wed, 6 Mar 2024 at 16:21, Palmer Dabbelt <palmer@dabbelt.com> wrote: > > > > > > > > On Wed, 06 Mar 2024 05:09:07 PST (-0800), Ard Biesheuvel wrote: > > > > > On Wed, 6 Mar 2024 at 14:02, Ard Biesheuvel <ardb@kernel.org> wrote: > > > > >> > > > > >> On Wed, 6 Mar 2024 at 13:34, yunhui cui <cuiyunhui@bytedance.com> wrote: > > > > >> > > > > > >> > Hi Ard, > > > > >> > > > > > >> > On Wed, Mar 6, 2024 at 5:36 PM Ard Biesheuvel <ardb@kernel.org> wrote: > > > > >> > > > > > > >> > > On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: > > > > >> > > > > > > > >> > > > Compared with gcc version 12, gcc version 13 uses the gp > > > > >> > > > register for compilation optimization, but the efistub module > > > > >> > > > does not initialize gp. > > > > >> > > > > > > > >> > > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> > > > > >> > > > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> > > > > >> > > > > > > >> > > This needs a sign-off, and your signoff needs to come after. > > > > >> > > > > > > >> > > > --- > > > > >> > > > arch/riscv/kernel/efi-header.S | 11 ++++++++++- > > > > >> > > > 1 file changed, 10 insertions(+), 1 deletion(-) > > > > >> > > > > > > > >> > > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S > > > > >> > > > index 515b2dfbca75..fa17c08c092a 100644 > > > > >> > > > --- a/arch/riscv/kernel/efi-header.S > > > > >> > > > +++ b/arch/riscv/kernel/efi-header.S > > > > >> > > > @@ -40,7 +40,7 @@ optional_header: > > > > >> > > > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData > > > > >> > > > #endif > > > > >> > > > .long 0 // SizeOfUninitializedData > > > > >> > > > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint > > > > >> > > > + .long _efistub_entry - _start // AddressOfEntryPoint > > > > >> > > > .long efi_header_end - _start // BaseOfCode > > > > >> > > > #ifdef CONFIG_32BIT > > > > >> > > > .long __pecoff_text_end - _start // BaseOfData > > > > >> > > > @@ -121,4 +121,13 @@ section_table: > > > > >> > > > > > > > >> > > > .balign 0x1000 > > > > >> > > > efi_header_end: > > > > >> > > > + > > > > >> > > > + .global _efistub_entry > > > > >> > > > +_efistub_entry: > > > > >> > > > > > > >> > > This should go into .text or .init.text, not the header. > > > > >> > > > > > > >> > > > + /* Reload the global pointer */ > > > > >> > > > + load_global_pointer > > > > >> > > > + > > > > >> > > > > > > >> > > What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI > > > > >> > > stub Makefile removes the SCS CFLAGS, so the stub will be built > > > > >> > > without shadow call stack support, which I guess means that it might > > > > >> > > use GP as a global pointer as usual? > > > > >> > > > > > > >> > > > + call __efistub_efi_pe_entry > > > > >> > > > + ret > > > > >> > > > + > > > > >> > > > > > > >> > > You are returning to the firmware here, but after modifying the GP > > > > >> > > register. Shouldn't you restore it to its old value? > > > > >> > There is no need to restore the value of the gp register. Where gp is > > > > >> > needed, the gp register must first be initialized. And here is the > > > > >> > entry. > > > > >> > > > > > >> > > > > >> But how should the firmware know that GP was corrupted after calling > > > > >> the kernel's EFI entrypoint? The EFI stub can return to the firmware > > > > >> if it encounters any errors while still running in the EFI boot > > > > >> services. > > > > >> > > > > > > > > > > Actually, I wonder if GP can be modified at all before > > > > > ExitBootServices(). The EFI timer interrupt is still live at this > > > > > point, and so the firmware is being called behind your back, and might > > > > > rely on GP retaining its original value. > > > > > > > > [A few of us are talking on IRC as I'm writing this...] > > > > > > > > The UEFI spec says "UEFI firmware must neither trust the > > > > values of tp and gp nor make an assumption of owning the write access to > > > > these register in any circumstances". It's kind of vague what "UEFI > > > > firmware" means here, but I think it's reasonable to assume that the > > > > kernel (and thus the EFI stub) is not included there. > > > > > > > > So under that interpretation, the kernel (including the EFI stub) would > > > > be allowed to overwrite GP with whatever it wants. > > > > > > > > > > OK, so even if the UEFI spec seems to suggest that using GP in EFI > > > applications such as the Linux EFI stub should be safe, I'd still like > > > to understand why this change is necessary. The patches you are > > > reverting are supposed to ensure that a) the compiler does not > > > generate references that can be relaxed to GP based ones, and b) no > > > R_RISCV_RELAX relocations are present in any of the code that runs in > > > the context of the EFI firmware. > > > > > > Are you still seeing GP based symbol references? Is there C code that > > > gets pulled into the EFI stub that uses GP based relocations perhaps? > > > (see list below). If any of those are implemented in C, they should > > > not be used by the EFI stub directly unless they are guaranteed to be > > > uninstrumented and callable at arbitrary offsets other than the one > > > they were linked to run at. > > > > > > > > > __efistub_memcmp = memcmp; > > > __efistub_memchr = memchr; > > > __efistub_memcpy = memcpy; > > > __efistub_memmove = memmove; > > > __efistub_memset = memset; > > > __efistub_strlen = strlen; > > > __efistub_strnlen = strnlen; > > > __efistub_strcmp = strcmp; > > > __efistub_strncmp = strncmp; > > > __efistub_strrchr = strrchr; > > > __efistub___memcpy = memcpy; > > > __efistub___memmove = memmove; > > > __efistub___memset = memset; > > > __efistub__start = _start; > > > __efistub__start_kernel = _start_kernel; > > > > > > (from arch/riscv/kernel/image-vars.h) > > > > Uhm never mind - these are all gone now, I was looking at a v6.1 > > kernel source tree. > > > > So that means that, as far as I can tell, the only kernel C code that > > executes in the context of the EFI firmware is built with -mno-relax > > and is checked for the absence of R_RISCV_RELAX relocations. So I fail > > to see why these changes are needed. > > > > Yunhui, could you please explain the reason for this series? > > From the logic of binutils, if "__global_pointer$" exists, it is > possible to use GP for optimization. For RISC-V, "__global_pointer$" > was introduced in commit "fbe934d69eb7e". Therefore, for the system as > a whole, we should keep using GP uniformly. There is no 'system as a whole' that can use GP 'uniformly' The EFI stub is a separate executable that runs from a different mapping of memory, in an execution context managed by the firmware. It happens to be linked into the same executable as the vmlinux kernel. > The root cause of this > problem is that GP is not loaded, rather than "On RISC-V, we also > avoid GP based relocations..." as commit "d2baf8cc82c17" said. GP is not loaded because in the EFI firmware context, there is no safe way to rely on it. > We need > to address problems head-on, rather than avoid them. > So what solution are you proposing for the potential GP conflicts between the boot loader, the Linux EFI stub and the firmware?
Hi Ard, On Fri, Mar 8, 2024 at 12:49 AM Ard Biesheuvel <ardb@kernel.org> wrote: > > On Thu, 7 Mar 2024 at 04:19, yunhui cui <cuiyunhui@bytedance.com> wrote: > > > > Hi Ard, > > > > On Thu, Mar 7, 2024 at 12:15 AM Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > > On Wed, 6 Mar 2024 at 16:44, Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > > > > On Wed, 6 Mar 2024 at 16:21, Palmer Dabbelt <palmer@dabbelt.com> wrote: > > > > > > > > > > On Wed, 06 Mar 2024 05:09:07 PST (-0800), Ard Biesheuvel wrote: > > > > > > On Wed, 6 Mar 2024 at 14:02, Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > >> > > > > > >> On Wed, 6 Mar 2024 at 13:34, yunhui cui <cuiyunhui@bytedance.com> wrote: > > > > > >> > > > > > > >> > Hi Ard, > > > > > >> > > > > > > >> > On Wed, Mar 6, 2024 at 5:36 PM Ard Biesheuvel <ardb@kernel.org> wrote: > > > > > >> > > > > > > > >> > > On Wed, 6 Mar 2024 at 09:56, Yunhui Cui <cuiyunhui@bytedance.com> wrote: > > > > > >> > > > > > > > > >> > > > Compared with gcc version 12, gcc version 13 uses the gp > > > > > >> > > > register for compilation optimization, but the efistub module > > > > > >> > > > does not initialize gp. > > > > > >> > > > > > > > > >> > > > Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> > > > > > >> > > > Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> > > > > > >> > > > > > > > >> > > This needs a sign-off, and your signoff needs to come after. > > > > > >> > > > > > > > >> > > > --- > > > > > >> > > > arch/riscv/kernel/efi-header.S | 11 ++++++++++- > > > > > >> > > > 1 file changed, 10 insertions(+), 1 deletion(-) > > > > > >> > > > > > > > > >> > > > diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S > > > > > >> > > > index 515b2dfbca75..fa17c08c092a 100644 > > > > > >> > > > --- a/arch/riscv/kernel/efi-header.S > > > > > >> > > > +++ b/arch/riscv/kernel/efi-header.S > > > > > >> > > > @@ -40,7 +40,7 @@ optional_header: > > > > > >> > > > .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData > > > > > >> > > > #endif > > > > > >> > > > .long 0 // SizeOfUninitializedData > > > > > >> > > > - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint > > > > > >> > > > + .long _efistub_entry - _start // AddressOfEntryPoint > > > > > >> > > > .long efi_header_end - _start // BaseOfCode > > > > > >> > > > #ifdef CONFIG_32BIT > > > > > >> > > > .long __pecoff_text_end - _start // BaseOfData > > > > > >> > > > @@ -121,4 +121,13 @@ section_table: > > > > > >> > > > > > > > > >> > > > .balign 0x1000 > > > > > >> > > > efi_header_end: > > > > > >> > > > + > > > > > >> > > > + .global _efistub_entry > > > > > >> > > > +_efistub_entry: > > > > > >> > > > > > > > >> > > This should go into .text or .init.text, not the header. > > > > > >> > > > > > > > >> > > > + /* Reload the global pointer */ > > > > > >> > > > + load_global_pointer > > > > > >> > > > + > > > > > >> > > > > > > > >> > > What is supposed to happen here if CONFIG_SHADOW_CALL_STACK=y? The EFI > > > > > >> > > stub Makefile removes the SCS CFLAGS, so the stub will be built > > > > > >> > > without shadow call stack support, which I guess means that it might > > > > > >> > > use GP as a global pointer as usual? > > > > > >> > > > > > > > >> > > > + call __efistub_efi_pe_entry > > > > > >> > > > + ret > > > > > >> > > > + > > > > > >> > > > > > > > >> > > You are returning to the firmware here, but after modifying the GP > > > > > >> > > register. Shouldn't you restore it to its old value? > > > > > >> > There is no need to restore the value of the gp register. Where gp is > > > > > >> > needed, the gp register must first be initialized. And here is the > > > > > >> > entry. > > > > > >> > > > > > > >> > > > > > >> But how should the firmware know that GP was corrupted after calling > > > > > >> the kernel's EFI entrypoint? The EFI stub can return to the firmware > > > > > >> if it encounters any errors while still running in the EFI boot > > > > > >> services. > > > > > >> > > > > > > > > > > > > Actually, I wonder if GP can be modified at all before > > > > > > ExitBootServices(). The EFI timer interrupt is still live at this > > > > > > point, and so the firmware is being called behind your back, and might > > > > > > rely on GP retaining its original value. > > > > > > > > > > [A few of us are talking on IRC as I'm writing this...] > > > > > > > > > > The UEFI spec says "UEFI firmware must neither trust the > > > > > values of tp and gp nor make an assumption of owning the write access to > > > > > these register in any circumstances". It's kind of vague what "UEFI > > > > > firmware" means here, but I think it's reasonable to assume that the > > > > > kernel (and thus the EFI stub) is not included there. > > > > > > > > > > So under that interpretation, the kernel (including the EFI stub) would > > > > > be allowed to overwrite GP with whatever it wants. > > > > > > > > > > > > > OK, so even if the UEFI spec seems to suggest that using GP in EFI > > > > applications such as the Linux EFI stub should be safe, I'd still like > > > > to understand why this change is necessary. The patches you are > > > > reverting are supposed to ensure that a) the compiler does not > > > > generate references that can be relaxed to GP based ones, and b) no > > > > R_RISCV_RELAX relocations are present in any of the code that runs in > > > > the context of the EFI firmware. > > > > > > > > Are you still seeing GP based symbol references? Is there C code that > > > > gets pulled into the EFI stub that uses GP based relocations perhaps? > > > > (see list below). If any of those are implemented in C, they should > > > > not be used by the EFI stub directly unless they are guaranteed to be > > > > uninstrumented and callable at arbitrary offsets other than the one > > > > they were linked to run at. > > > > > > > > > > > > __efistub_memcmp = memcmp; > > > > __efistub_memchr = memchr; > > > > __efistub_memcpy = memcpy; > > > > __efistub_memmove = memmove; > > > > __efistub_memset = memset; > > > > __efistub_strlen = strlen; > > > > __efistub_strnlen = strnlen; > > > > __efistub_strcmp = strcmp; > > > > __efistub_strncmp = strncmp; > > > > __efistub_strrchr = strrchr; > > > > __efistub___memcpy = memcpy; > > > > __efistub___memmove = memmove; > > > > __efistub___memset = memset; > > > > __efistub__start = _start; > > > > __efistub__start_kernel = _start_kernel; > > > > > > > > (from arch/riscv/kernel/image-vars.h) > > > > > > Uhm never mind - these are all gone now, I was looking at a v6.1 > > > kernel source tree. > > > > > > So that means that, as far as I can tell, the only kernel C code that > > > executes in the context of the EFI firmware is built with -mno-relax > > > and is checked for the absence of R_RISCV_RELAX relocations. So I fail > > > to see why these changes are needed. > > > > > > Yunhui, could you please explain the reason for this series? > > > > From the logic of binutils, if "__global_pointer$" exists, it is > > possible to use GP for optimization. For RISC-V, "__global_pointer$" > > was introduced in commit "fbe934d69eb7e". Therefore, for the system as > > a whole, we should keep using GP uniformly. > > There is no 'system as a whole' that can use GP 'uniformly' > > The EFI stub is a separate executable that runs from a different > mapping of memory, in an execution context managed by the firmware. It > happens to be linked into the same executable as the vmlinux kernel. > > > The root cause of this > > problem is that GP is not loaded, rather than "On RISC-V, we also > > avoid GP based relocations..." as commit "d2baf8cc82c17" said. > > GP is not loaded because in the EFI firmware context, there is no safe > way to rely on it. > > > We need > > to address problems head-on, rather than avoid them. > > > > So what solution are you proposing for the potential GP conflicts > between the boot loader, the Linux EFI stub and the firmware? The GP register values are now loaded in the arch/riscv/kernel/head.S and arch/riscv/kernel/suspend_entry.S files. Let's think about EFI runtimeservice. If the EFI firmware code uses GP registers but the compiler does not avoid GP, and kernel uses the callback function provided by EFI, is there a problem? Is it possible to solve the problem only by making the firmware code not use GP at all and compiling options to avoid using GP? The same goes for efistub. So the way to solve this problem is that the firmware does not use GP optimization. Does this allow efistub to load the GP register? Thanks, Yunhui
On Fri, 8 Mar 2024 at 08:10, yunhui cui <cuiyunhui@bytedance.com> wrote: > > Hi Ard, > > On Fri, Mar 8, 2024 at 12:49 AM Ard Biesheuvel <ardb@kernel.org> wrote: > > > > On Thu, 7 Mar 2024 at 04:19, yunhui cui <cuiyunhui@bytedance.com> wrote: > > > ... > > > > > > From the logic of binutils, if "__global_pointer$" exists, it is > > > possible to use GP for optimization. For RISC-V, "__global_pointer$" > > > was introduced in commit "fbe934d69eb7e". Therefore, for the system as > > > a whole, we should keep using GP uniformly. > > > > There is no 'system as a whole' that can use GP 'uniformly' > > > > The EFI stub is a separate executable that runs from a different > > mapping of memory, in an execution context managed by the firmware. It > > happens to be linked into the same executable as the vmlinux kernel. > > > > > The root cause of this > > > problem is that GP is not loaded, rather than "On RISC-V, we also > > > avoid GP based relocations..." as commit "d2baf8cc82c17" said. > > > > GP is not loaded because in the EFI firmware context, there is no safe > > way to rely on it. > > > > > We need > > > to address problems head-on, rather than avoid them. > > > > > > > So what solution are you proposing for the potential GP conflicts > > between the boot loader, the Linux EFI stub and the firmware? > > > The GP register values are now loaded in the arch/riscv/kernel/head.S > and arch/riscv/kernel/suspend_entry.S files. > > Let's think about EFI runtimeservice. If the EFI firmware code uses GP > registers but the compiler does not avoid GP, and kernel uses the > callback function provided by EFI, is there a problem? Is it possible > to solve the problem only by making the firmware code not use GP at > all and compiling options to avoid using GP? > EFI runtime services do not use callbacks, and execute in a context that is entirely owned by the OS. So this is one place where EFI firmware cannot use GP at all even if the UEFI spec permitted it. > The same goes for efistub. > Not really. The UEFI spec seems to suggest that *system* firmware should not touch GP or make any assumptions about its value, but it doesn't say anything about EFI applications such as the EFI stub or GRUB. > So the way to solve this problem is that the firmware does not use GP > optimization. Does this allow efistub to load the GP register? > What about GRUB or other bootloaders that are loaded before the kernel, but are still active while the EFI stub is executing? Who gets to own GP in this scenario?
diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-header.S index 515b2dfbca75..fa17c08c092a 100644 --- a/arch/riscv/kernel/efi-header.S +++ b/arch/riscv/kernel/efi-header.S @@ -40,7 +40,7 @@ optional_header: .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedData #endif .long 0 // SizeOfUninitializedData - .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint + .long _efistub_entry - _start // AddressOfEntryPoint .long efi_header_end - _start // BaseOfCode #ifdef CONFIG_32BIT .long __pecoff_text_end - _start // BaseOfData @@ -121,4 +121,13 @@ section_table: .balign 0x1000 efi_header_end: + + .global _efistub_entry +_efistub_entry: + /* Reload the global pointer */ + load_global_pointer + + call __efistub_efi_pe_entry + ret + .endm
Compared with gcc version 12, gcc version 13 uses the gp register for compilation optimization, but the efistub module does not initialize gp. Signed-off-by: Yunhui Cui <cuiyunhui@bytedance.com> Co-Developed-by: Zhipeng Xu <xuzhipeng.1973@bytedance.com> --- arch/riscv/kernel/efi-header.S | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)