From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from smtp.kernel.org (aws-us-west-2-korg-mail-1.web.codeaurora.org [10.30.226.201]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id E4D166125 for ; Fri, 24 Feb 2023 15:58:55 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id 118FDC433D2; Fri, 24 Feb 2023 15:58:54 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1677254335; bh=rRfXjsGBMvvnDz60uwLB5dc9ju4oNbQeuN0GLhjFFZk=; h=From:To:Subject:In-Reply-To:References:Date:From; b=IV+x2SOKujlDxRkpOOclYLhRMxavxs2ZgAVNSvw7w90V9xvyuyFfsJCv80gYcOG2p A4TsrQ/qWG2iKNVl3avpevUIt5GA93ovI/qxOn7KRfjQxoOnikKjPiXeR7hhdclS6q dqmICBshEsrCnUZnXi2fTr77wbblb1vHL/rZvDQG8+JRWHjmvTQn465vrIBEOHJ7Iz OcdwuhdpCS1R/LJKz0LToyAjRK19HbZPcprjDc1e+B3ESMEvkoWdKClgQQu+6mliia B68A719Xz9UxYpvtgkDRGcfVlkhGZKsdVyehclJsK+pzSytz6pbulYjyBSnzYllZVj F5Yvnz7APGcSQ== From: =?utf-8?B?QmrDtnJuIFTDtnBlbA==?= To: Alexandre Ghiti , Alexandre Ghiti , Michael Ellerman , Nicholas Piggin , Christophe Leroy , Paul Walmsley , Palmer Dabbelt , Albert Ou , linuxppc-dev@lists.ozlabs.org, linux-kernel@vger.kernel.org, linux-riscv@lists.infradead.org, nathan@kernel.org, linux-kbuild@vger.kernel.org, llvm@lists.linux.dev, ndesaulniers@google.com Subject: Re: [PATCH v8 1/3] riscv: Introduce CONFIG_RELOCATABLE In-Reply-To: <4a6fc7a3-9697-a49b-0941-97f32194b0d7@ghiti.fr> References: <20230215143626.453491-1-alexghiti@rivosinc.com> <20230215143626.453491-2-alexghiti@rivosinc.com> <4a6fc7a3-9697-a49b-0941-97f32194b0d7@ghiti.fr> Date: Fri, 24 Feb 2023 16:58:52 +0100 Message-ID: <877cw7dphf.fsf@all.your.base.are.belong.to.us> Precedence: bulk X-Mailing-List: llvm@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Alexandre Ghiti writes: > +cc linux-kbuild, llvm, Nathan, Nick > > On 2/15/23 15:36, Alexandre Ghiti wrote: >> From: Alexandre Ghiti >> >> This config allows to compile 64b kernel as PIE and to relocate it at >> any virtual address at runtime: this paves the way to KASLR. >> Runtime relocation is possible since relocation metadata are embedded in= to >> the kernel. >> >> Note that relocating at runtime introduces an overhead even if the >> kernel is loaded at the same address it was linked at and that the compi= ler >> options are those used in arm64 which uses the same RELA relocation >> format. >> >> Signed-off-by: Alexandre Ghiti >> --- >> arch/riscv/Kconfig | 14 +++++++++ >> arch/riscv/Makefile | 7 +++-- >> arch/riscv/kernel/efi-header.S | 6 ++-- >> arch/riscv/kernel/vmlinux.lds.S | 10 ++++-- >> arch/riscv/mm/Makefile | 4 +++ >> arch/riscv/mm/init.c | 54 ++++++++++++++++++++++++++++++++- >> 6 files changed, 87 insertions(+), 8 deletions(-) >> >> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig >> index e2b656043abf..e0ee7ce4b2e3 100644 >> --- a/arch/riscv/Kconfig >> +++ b/arch/riscv/Kconfig >> @@ -544,6 +544,20 @@ config COMPAT >>=20=20=20 >> If you want to execute 32-bit userspace applications, say Y. >>=20=20=20 >> +config RELOCATABLE >> + bool "Build a relocatable kernel" >> + depends on MMU && 64BIT && !XIP_KERNEL >> + help >> + This builds a kernel as a Position Independent Executable (PI= E), >> + which retains all relocation metadata required to relocate the >> + kernel binary at runtime to a different virtual address than = the >> + address it was linked at. >> + Since RISCV uses the RELA relocation format, this requires a >> + relocation pass at runtime even if the kernel is loaded at the >> + same address it was linked at. >> + >> + If unsure, say N. >> + >> endmenu # "Kernel features" >>=20=20=20 >> menu "Boot options" >> diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile >> index 82153960ac00..97c34136b027 100644 >> --- a/arch/riscv/Makefile >> +++ b/arch/riscv/Makefile >> @@ -7,9 +7,12 @@ >> # >>=20=20=20 >> OBJCOPYFLAGS :=3D -O binary >> -LDFLAGS_vmlinux :=3D >> +ifeq ($(CONFIG_RELOCATABLE),y) >> + LDFLAGS_vmlinux +=3D -shared -Bsymbolic -z notext -z norelro >> + KBUILD_CFLAGS +=3D -fPIE >> +endif >> ifeq ($(CONFIG_DYNAMIC_FTRACE),y) >> - LDFLAGS_vmlinux :=3D --no-relax >> + LDFLAGS_vmlinux +=3D --no-relax >> KBUILD_CPPFLAGS +=3D -DCC_USING_PATCHABLE_FUNCTION_ENTRY >> CC_FLAGS_FTRACE :=3D -fpatchable-function-entry=3D8 >> endif >> diff --git a/arch/riscv/kernel/efi-header.S b/arch/riscv/kernel/efi-head= er.S >> index 8e733aa48ba6..f7ee09c4f12d 100644 >> --- a/arch/riscv/kernel/efi-header.S >> +++ b/arch/riscv/kernel/efi-header.S >> @@ -33,7 +33,7 @@ optional_header: >> .byte 0x02 // MajorLinkerVersion >> .byte 0x14 // MinorLinkerVersion >> .long __pecoff_text_end - efi_header_end // SizeOfCode >> - .long __pecoff_data_virt_size // SizeOfInitializedData >> + .long __pecoff_data_virt_end - __pecoff_text_end // SizeOfInitializedD= ata >> .long 0 // SizeOfUninitializedData >> .long __efistub_efi_pe_entry - _start // AddressOfEntryPoint >> .long efi_header_end - _start // BaseOfCode >> @@ -91,9 +91,9 @@ section_table: >> IMAGE_SCN_MEM_EXECUTE // Characteristics >>=20=20=20 >> .ascii ".data\0\0\0" >> - .long __pecoff_data_virt_size // VirtualSize >> + .long __pecoff_data_virt_end - __pecoff_text_end // VirtualSize >> .long __pecoff_text_end - _start // VirtualAddress >> - .long __pecoff_data_raw_size // SizeOfRawData >> + .long __pecoff_data_raw_end - __pecoff_text_end // SizeOfRawData >> .long __pecoff_text_end - _start // PointerToRawData >>=20=20=20 >> .long 0 // PointerToRelocations >> diff --git a/arch/riscv/kernel/vmlinux.lds.S b/arch/riscv/kernel/vmlinux= .lds.S >> index 4e6c88aa4d87..8be2de3be08c 100644 >> --- a/arch/riscv/kernel/vmlinux.lds.S >> +++ b/arch/riscv/kernel/vmlinux.lds.S >> @@ -122,9 +122,15 @@ SECTIONS >> *(.sdata*) >> } >>=20=20=20 >> + .rela.dyn : ALIGN(8) { >> + __rela_dyn_start =3D .; >> + *(.rela .rela*) >> + __rela_dyn_end =3D .; >> + } >> + > > > So I realized those relocations would be better in the init section so=20 > we can get rid of them at some point. So I tried the following: > > diff --git a/arch/riscv/kernel/vmlinux.lds.S=20 > b/arch/riscv/kernel/vmlinux.lds.S > index 7ac215467fd5..6111023a89ef 100644 > --- a/arch/riscv/kernel/vmlinux.lds.S > +++ b/arch/riscv/kernel/vmlinux.lds.S > @@ -93,6 +93,12 @@ SECTIONS > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0= =C2=A0=C2=A0=C2=A0 *(.rel.dyn*) > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 } > > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 .rela.dyn : ALIGN(8) { > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0= =C2=A0=C2=A0 __rela_dyn_start =3D .; > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0= =C2=A0=C2=A0 *(.rela .rela*) > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0= =C2=A0=C2=A0 __rela_dyn_end =3D .; > +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 } > + > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 __init_data_end =3D .; > > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 . =3D ALIGN(8); > @@ -119,12 +125,6 @@ SECTIONS > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0= =C2=A0=C2=A0=C2=A0 *(.sdata*) > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 } > > -=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 .rela.dyn : ALIGN(8) { > -=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0= =C2=A0=C2=A0 __rela_dyn_start =3D .; > -=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0= =C2=A0=C2=A0 *(.rela .rela*) > -=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0= =C2=A0=C2=A0 __rela_dyn_end =3D .; > -=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 } > - > =C2=A0#ifdef CONFIG_EFI > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 .pecoff_edata_padding : { BYT= E(0); . =3D=20 > ALIGN(PECOFF_FILE_ALIGNMENT); } > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 __pecoff_data_raw_end =3D ABS= OLUTE(.); > > > But then all the relocations in vmlinux end up being null: > > vmlinux:=C2=A0=C2=A0=C2=A0=C2=A0 file format elf64-littleriscv > > $ riscv64-linux-gnu-objdump -R vmlinux > > DYNAMIC RELOCATION RECORDS > OFFSET=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 TYPE= =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2= =A0 VALUE > 0000000000000000 R_RISCV_NONE=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 *ABS* > 0000000000000000 R_RISCV_NONE=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 *ABS* > .... > > =C2=A0I also noticed that re-linking vmlinux with the same command right= =20 > after works (ie, the relocations are now valid): > > $ riscv64-linux-gnu-objdump -R vmlinux > > vmlinux:=C2=A0=C2=A0=C2=A0=C2=A0 file format elf64-littleriscv > > DYNAMIC RELOCATION RECORDS > OFFSET=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 TYPE= =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2= =A0 VALUE > ffffffff82600718 R_RISCV_RELATIVE=C2=A0 *ABS*-0x000000007d9ff8e8 > ffffffff82600720 R_RISCV_RELATIVE=C2=A0 *ABS*-0x000000007d9ff8e8 > ... > > Below is the command used to generate this working vmlinux: > > riscv64-unknown-linux-gnu-ld -melf64lriscv -z noexecstack=20 > --no-warn-rwx-segments -shared -Bsymbolic -z notext -z norelro=20 > --no-relax --build-id=3Dsha1 --script=3D./arch/riscv/kernel/vmlinux.lds=20 > -Map=3Dvmlinux.map -o vmlinux --whole-archive vmlinux.a .vmlinux.export.o= =20 > init/version-timestamp.o --no-whole-archive --start-group=20 > ./drivers/firmware/efi/libstub/lib.a --end-group .tmp_vmlinux.kallsyms3.o > > I tried a lot of things, but I struggle to understand, does anyone have=20 > any idea? FYI, the same problem happens with LLVM. Don't ask me *why*, but adding --emit-relocs to your linker flags solves "the NULL .rela.dyn" both for GCC and LLVM. The downside is that you end up with a bunch of .rela cruft in your vmlinux. Bj=C3=B6rn