diff mbox series

[v5,2/4] riscv: Introduce CONFIG_RELOCATABLE

Message ID 20200607075949.665-3-alex@ghiti.fr (mailing list archive)
State New, archived
Headers show
Series vmalloc kernel mapping and relocatable kernel | expand

Commit Message

Alexandre Ghiti June 7, 2020, 7:59 a.m. UTC
This config allows to compile the kernel as PIE and to relocate it at
any virtual address at runtime: this paves the way to KASLR and to 4-level
page table folding at runtime. Runtime relocation is possible since
relocation metadata are embedded into 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 compiler
options are those used in arm64 which uses the same RELA relocation
format.

Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Reviewed-by: Zong Li <zong.li@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
---
 arch/riscv/Kconfig              | 12 +++++++
 arch/riscv/Makefile             |  5 ++-
 arch/riscv/kernel/vmlinux.lds.S |  6 ++--
 arch/riscv/mm/Makefile          |  4 +++
 arch/riscv/mm/init.c            | 63 +++++++++++++++++++++++++++++++++
 5 files changed, 87 insertions(+), 3 deletions(-)

Comments

Jerome Forissier June 10, 2020, 2:10 p.m. UTC | #1
On 6/7/20 9:59 AM, Alexandre Ghiti wrote:
[...]

> +config RELOCATABLE
> +	bool
> +	depends on MMU
> +	help
> +          This builds a kernel as a Position Independent Executable (PIE),
> +          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.

Is this true? I thought that the GNU linker would write the "proper"
values by default, contrary to the LLVM linker (ld.lld) which would need
a special flag: --apply-dynamic-relocs (by default the relocated places
are set to zero). At least, it is my experience with Aarch64 on a
different project. So, sorry if I'm talking nonsense here -- I have not
looked at the details.
Alexandre Ghiti June 11, 2020, 7:43 p.m. UTC | #2
Hi Jerome,

Le 6/10/20 à 10:10 AM, Jerome Forissier a écrit :
> On 6/7/20 9:59 AM, Alexandre Ghiti wrote:
> [...]
>
>> +config RELOCATABLE
>> +	bool
>> +	depends on MMU
>> +	help
>> +          This builds a kernel as a Position Independent Executable (PIE),
>> +          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.
> Is this true? I thought that the GNU linker would write the "proper"
> values by default, contrary to the LLVM linker (ld.lld) which would need
> a special flag: --apply-dynamic-relocs (by default the relocated places
> are set to zero). At least, it is my experience with Aarch64 on a
> different project. So, sorry if I'm talking nonsense here -- I have not
> looked at the details.
>
>

It seems that you're right, at least for aarch64 since they specifically 
specify the --no-apply-dynamic-relocs option. I retried to boot without 
relocating at runtime, and it fails on riscv. Can this be arch specific ?

Thanks,

Alex
diff mbox series

Patch

diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index a31e1a41913a..93127d5913fe 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -170,6 +170,18 @@  config PGTABLE_LEVELS
 	default 3 if 64BIT
 	default 2
 
+config RELOCATABLE
+	bool
+	depends on MMU
+	help
+          This builds a kernel as a Position Independent Executable (PIE),
+          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.
+
 source "arch/riscv/Kconfig.socs"
 
 menu "Platform type"
diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
index fb6e37db836d..1406416ea743 100644
--- a/arch/riscv/Makefile
+++ b/arch/riscv/Makefile
@@ -9,7 +9,10 @@ 
 #
 
 OBJCOPYFLAGS    := -O binary
-LDFLAGS_vmlinux :=
+ifeq ($(CONFIG_RELOCATABLE),y)
+LDFLAGS_vmlinux := -shared -Bsymbolic -z notext -z norelro
+KBUILD_CFLAGS += -fPIE
+endif
 ifeq ($(CONFIG_DYNAMIC_FTRACE),y)
 	LDFLAGS_vmlinux := --no-relax
 endif
diff --git a/arch/riscv/kernel/vmlinux.lds.S b/arch/riscv/kernel/vmlinux.lds.S
index a9abde62909f..e8ffba8c2044 100644
--- a/arch/riscv/kernel/vmlinux.lds.S
+++ b/arch/riscv/kernel/vmlinux.lds.S
@@ -85,8 +85,10 @@  SECTIONS
 
 	BSS_SECTION(PAGE_SIZE, PAGE_SIZE, 0)
 
-	.rel.dyn : {
-		*(.rel.dyn*)
+	.rela.dyn : ALIGN(8) {
+		__rela_dyn_start = .;
+		*(.rela .rela*)
+		__rela_dyn_end = .;
 	}
 
 	_end = .;
diff --git a/arch/riscv/mm/Makefile b/arch/riscv/mm/Makefile
index 363ef01c30b1..dc5cdaa80bc1 100644
--- a/arch/riscv/mm/Makefile
+++ b/arch/riscv/mm/Makefile
@@ -1,6 +1,10 @@ 
 # SPDX-License-Identifier: GPL-2.0-only
 
 CFLAGS_init.o := -mcmodel=medany
+ifdef CONFIG_RELOCATABLE
+CFLAGS_init.o += -fno-pie
+endif
+
 ifdef CONFIG_FTRACE
 CFLAGS_REMOVE_init.o = -pg
 endif
diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c
index 71da78914645..29b33289a12f 100644
--- a/arch/riscv/mm/init.c
+++ b/arch/riscv/mm/init.c
@@ -13,6 +13,9 @@ 
 #include <linux/of_fdt.h>
 #include <linux/libfdt.h>
 #include <linux/set_memory.h>
+#ifdef CONFIG_RELOCATABLE
+#include <linux/elf.h>
+#endif
 
 #include <asm/fixmap.h>
 #include <asm/tlbflush.h>
@@ -379,6 +382,53 @@  static uintptr_t __init best_map_size(phys_addr_t base, phys_addr_t size)
 #error "setup_vm() is called from head.S before relocate so it should not use absolute addressing."
 #endif
 
+#ifdef CONFIG_RELOCATABLE
+extern unsigned long __rela_dyn_start, __rela_dyn_end;
+
+#ifdef CONFIG_64BIT
+#define Elf_Rela Elf64_Rela
+#define Elf_Addr Elf64_Addr
+#else
+#define Elf_Rela Elf32_Rela
+#define Elf_Addr Elf32_Addr
+#endif
+
+void __init relocate_kernel(uintptr_t load_pa)
+{
+	Elf_Rela *rela = (Elf_Rela *)&__rela_dyn_start;
+	/*
+	 * This holds the offset between the linked virtual address and the
+	 * relocated virtual address.
+	 */
+	uintptr_t reloc_offset = kernel_virt_addr - KERNEL_LINK_ADDR;
+	/*
+	 * This holds the offset between kernel linked virtual address and
+	 * physical address.
+	 */
+	uintptr_t va_kernel_link_pa_offset = KERNEL_LINK_ADDR - load_pa;
+
+	for ( ; rela < (Elf_Rela *)&__rela_dyn_end; rela++) {
+		Elf_Addr addr = (rela->r_offset - va_kernel_link_pa_offset);
+		Elf_Addr relocated_addr = rela->r_addend;
+
+		if (rela->r_info != R_RISCV_RELATIVE)
+			continue;
+
+		/*
+		 * Make sure to not relocate vdso symbols like rt_sigreturn
+		 * which are linked from the address 0 in vmlinux since
+		 * vdso symbol addresses are actually used as an offset from
+		 * mm->context.vdso in VDSO_OFFSET macro.
+		 */
+		if (relocated_addr >= KERNEL_LINK_ADDR)
+			relocated_addr += reloc_offset;
+
+		*(Elf_Addr *)addr = relocated_addr;
+	}
+}
+
+#endif
+
 static uintptr_t load_pa, load_sz;
 
 static void __init create_kernel_page_table(pgd_t *pgdir, uintptr_t map_size)
@@ -405,6 +455,19 @@  asmlinkage void __init setup_vm(uintptr_t dtb_pa)
 
 	pfn_base = PFN_DOWN(load_pa);
 
+#ifdef CONFIG_RELOCATABLE
+#ifdef CONFIG_64BIT
+	/*
+	 * Early page table uses only one PGDIR, which makes it possible
+	 * to map PGDIR_SIZE aligned on PGDIR_SIZE: if the relocation offset
+	 * makes the kernel cross over a PGDIR_SIZE boundary, raise a bug
+	 * since a part of the kernel would not get mapped.
+	 * This cannot happen on rv32 as we use the entire page directory level.
+	 */
+	BUG_ON(PGDIR_SIZE - (kernel_virt_addr & (PGDIR_SIZE - 1)) < load_sz);
+#endif
+	relocate_kernel(load_pa);
+#endif
 	/*
 	 * Enforce boot alignment requirements of RV32 and
 	 * RV64 by only allowing PMD or PGD mappings.