The following KASLR implementation allows to randomize the kernel mapping:
- virtually: we expect the bootloader to provide a seed in the device-tree
- physically: only implemented in the EFI stub, it relies on the firmware to
provide a seed using EFI_RNG_PROTOCOL. arm64 has a similar implementation
hence the patch 3 factorizes KASLR related functions for riscv to take
advantage.
The new virtual kernel location is limited by the early page table that only
has one PUD and with the PMD alignment constraint, the kernel can only take
< 512 positions.
base-commit-tag: v6.5-rc1
Changes in v5:
* Renamed efi-stub-kaslr.c into kaslr.c and fix commit log of patch 3,
as suggested by Ard
* Removed stubs since the kaslr functions were moved to their own file
(and then does not trigger any build failure for architectures that do
not call those functions since they are in their own compilation unit)
Changes in v4:
* Fix efi_get_kimg macro that returned nothing
* Moved new kaslr functions into their own files to avoid zboot link
failures, as suggested by Ard
Changes in v3:
* Rebase on top of 6.4-rc2
* Make RANDOMIZE_BASE depend on 64bit
* Fix efi_icache_sync and efi_get_kimg_min_align which were undefined
in x86 (and certainly other archs)
* Add patch 4 to fix warning on rv32
Changes in v2:
* Rebase on top of 6.3-rc1
* Add a riscv cache sync after memcpying the kernel
* Add kaslr_offset implementation for KCOV
* Add forward declaration to quiet LLVM
Alexandre Ghiti (5):
riscv: Introduce virtual kernel mapping KASLR
riscv: Dump out kernel offset information on panic
arm64: libstub: Move KASLR handling functions to kaslr.c
libstub: Fix compilation warning for rv32
riscv: libstub: Implement KASLR by using generic functions
arch/arm64/include/asm/efi.h | 2 +
arch/riscv/Kconfig | 19 +++
arch/riscv/include/asm/efi.h | 2 +
arch/riscv/include/asm/page.h | 3 +
arch/riscv/kernel/image-vars.h | 1 +
arch/riscv/kernel/pi/Makefile | 2 +-
arch/riscv/kernel/pi/cmdline_early.c | 13 ++
arch/riscv/kernel/pi/fdt_early.c | 30 ++++
arch/riscv/kernel/setup.c | 25 ++++
arch/riscv/mm/init.c | 36 ++++-
drivers/firmware/efi/libstub/Makefile | 3 +-
drivers/firmware/efi/libstub/arm64-stub.c | 117 ++--------------
drivers/firmware/efi/libstub/efistub.h | 8 ++
drivers/firmware/efi/libstub/kaslr.c | 159 ++++++++++++++++++++++
drivers/firmware/efi/libstub/riscv-stub.c | 33 ++---
15 files changed, 328 insertions(+), 125 deletions(-)
create mode 100644 arch/riscv/kernel/pi/fdt_early.c
create mode 100644 drivers/firmware/efi/libstub/kaslr.c
--
2.39.2
KASLR implementation relies on a relocatable kernel so that we can move
the kernel mapping.
The seed needed to virtually move the kernel is taken from the device tree,
so we rely on the bootloader to provide a correct seed. Zkr could be used
unconditionnally instead if implemented, but that's for another patch.
Signed-off-by: Alexandre Ghiti <[email protected]>
---
arch/riscv/Kconfig | 19 +++++++++++++++
arch/riscv/include/asm/page.h | 3 +++
arch/riscv/kernel/pi/Makefile | 2 +-
arch/riscv/kernel/pi/cmdline_early.c | 13 ++++++++++
arch/riscv/kernel/pi/fdt_early.c | 30 +++++++++++++++++++++++
arch/riscv/mm/init.c | 36 +++++++++++++++++++++++++++-
6 files changed, 101 insertions(+), 2 deletions(-)
create mode 100644 arch/riscv/kernel/pi/fdt_early.c
diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index 4c07b9189c86..6a606d5b74c6 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -719,6 +719,25 @@ config RELOCATABLE
If unsure, say N.
+config RANDOMIZE_BASE
+ bool "Randomize the address of the kernel image"
+ select RELOCATABLE
+ depends on MMU && 64BIT && !XIP_KERNEL
+ help
+ Randomizes the virtual address at which the kernel image is
+ loaded, as a security feature that deters exploit attempts
+ relying on knowledge of the location of kernel internals.
+
+ It is the bootloader's job to provide entropy, by passing a
+ random u64 value in /chosen/kaslr-seed at kernel entry.
+
+ When booting via the UEFI stub, it will invoke the firmware's
+ EFI_RNG_PROTOCOL implementation (if available) to supply entropy
+ to the kernel proper. In addition, it will randomise the physical
+ location of the kernel Image as well.
+
+ If unsure, say N.
+
endmenu # "Kernel features"
menu "Boot options"
diff --git a/arch/riscv/include/asm/page.h b/arch/riscv/include/asm/page.h
index b55ba20903ec..5488ecc337b6 100644
--- a/arch/riscv/include/asm/page.h
+++ b/arch/riscv/include/asm/page.h
@@ -106,6 +106,7 @@ typedef struct page *pgtable_t;
struct kernel_mapping {
unsigned long page_offset;
unsigned long virt_addr;
+ unsigned long virt_offset;
uintptr_t phys_addr;
uintptr_t size;
/* Offset between linear mapping virtual address and kernel load address */
@@ -185,6 +186,8 @@ extern phys_addr_t __phys_addr_symbol(unsigned long x);
#define sym_to_pfn(x) __phys_to_pfn(__pa_symbol(x))
+unsigned long kaslr_offset(void);
+
#endif /* __ASSEMBLY__ */
#define virt_addr_valid(vaddr) ({ \
diff --git a/arch/riscv/kernel/pi/Makefile b/arch/riscv/kernel/pi/Makefile
index 7b593d44c712..07915dc9279e 100644
--- a/arch/riscv/kernel/pi/Makefile
+++ b/arch/riscv/kernel/pi/Makefile
@@ -35,5 +35,5 @@ $(obj)/string.o: $(srctree)/lib/string.c FORCE
$(obj)/ctype.o: $(srctree)/lib/ctype.c FORCE
$(call if_changed_rule,cc_o_c)
-obj-y := cmdline_early.pi.o string.pi.o ctype.pi.o lib-fdt.pi.o lib-fdt_ro.pi.o
+obj-y := cmdline_early.pi.o fdt_early.pi.o string.pi.o ctype.pi.o lib-fdt.pi.o lib-fdt_ro.pi.o
extra-y := $(patsubst %.pi.o,%.o,$(obj-y))
diff --git a/arch/riscv/kernel/pi/cmdline_early.c b/arch/riscv/kernel/pi/cmdline_early.c
index 05652d13c746..68e786c84c94 100644
--- a/arch/riscv/kernel/pi/cmdline_early.c
+++ b/arch/riscv/kernel/pi/cmdline_early.c
@@ -14,6 +14,7 @@ static char early_cmdline[COMMAND_LINE_SIZE];
* LLVM complain because the function is actually unused in this file).
*/
u64 set_satp_mode_from_cmdline(uintptr_t dtb_pa);
+bool set_nokaslr_from_cmdline(uintptr_t dtb_pa);
static char *get_early_cmdline(uintptr_t dtb_pa)
{
@@ -60,3 +61,15 @@ u64 set_satp_mode_from_cmdline(uintptr_t dtb_pa)
return match_noXlvl(cmdline);
}
+
+static bool match_nokaslr(char *cmdline)
+{
+ return strstr(cmdline, "nokaslr");
+}
+
+bool set_nokaslr_from_cmdline(uintptr_t dtb_pa)
+{
+ char *cmdline = get_early_cmdline(dtb_pa);
+
+ return match_nokaslr(cmdline);
+}
diff --git a/arch/riscv/kernel/pi/fdt_early.c b/arch/riscv/kernel/pi/fdt_early.c
new file mode 100644
index 000000000000..899610e042ab
--- /dev/null
+++ b/arch/riscv/kernel/pi/fdt_early.c
@@ -0,0 +1,30 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/types.h>
+#include <linux/init.h>
+#include <linux/libfdt.h>
+
+/*
+ * Declare the functions that are exported (but prefixed) here so that LLVM
+ * does not complain it lacks the 'static' keyword (which, if added, makes
+ * LLVM complain because the function is actually unused in this file).
+ */
+u64 get_kaslr_seed(uintptr_t dtb_pa);
+
+u64 get_kaslr_seed(uintptr_t dtb_pa)
+{
+ int node, len;
+ fdt64_t *prop;
+ u64 ret;
+
+ node = fdt_path_offset((void *)dtb_pa, "/chosen");
+ if (node < 0)
+ return 0;
+
+ prop = fdt_getprop_w((void *)dtb_pa, node, "kaslr-seed", &len);
+ if (!prop || len != sizeof(u64))
+ return 0;
+
+ ret = fdt64_to_cpu(*prop);
+ *prop = 0;
+ return ret;
+}
diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c
index 70fb31960b63..ff926531236e 100644
--- a/arch/riscv/mm/init.c
+++ b/arch/riscv/mm/init.c
@@ -1012,11 +1012,45 @@ static void __init pt_ops_set_late(void)
#endif
}
+#ifdef CONFIG_RANDOMIZE_BASE
+extern bool __init __pi_set_nokaslr_from_cmdline(uintptr_t dtb_pa);
+extern u64 __init __pi_get_kaslr_seed(uintptr_t dtb_pa);
+
+static int __init print_nokaslr(char *p)
+{
+ pr_info("Disabled KASLR");
+ return 0;
+}
+early_param("nokaslr", print_nokaslr);
+
+unsigned long kaslr_offset(void)
+{
+ return kernel_map.virt_offset;
+}
+#endif
+
asmlinkage void __init setup_vm(uintptr_t dtb_pa)
{
pmd_t __maybe_unused fix_bmap_spmd, fix_bmap_epmd;
- kernel_map.virt_addr = KERNEL_LINK_ADDR;
+#ifdef CONFIG_RANDOMIZE_BASE
+ if (!__pi_set_nokaslr_from_cmdline(dtb_pa)) {
+ u64 kaslr_seed = __pi_get_kaslr_seed(dtb_pa);
+ u32 kernel_size = (uintptr_t)(&_end) - (uintptr_t)(&_start);
+ u32 nr_pos;
+
+ /*
+ * Compute the number of positions available: we are limited
+ * by the early page table that only has one PUD and we must
+ * be aligned on PMD_SIZE.
+ */
+ nr_pos = (PUD_SIZE - kernel_size) / PMD_SIZE;
+
+ kernel_map.virt_offset = (kaslr_seed % nr_pos) * PMD_SIZE;
+ }
+#endif
+
+ kernel_map.virt_addr = KERNEL_LINK_ADDR + kernel_map.virt_offset;
kernel_map.page_offset = _AC(CONFIG_PAGE_OFFSET, UL);
#ifdef CONFIG_XIP_KERNEL
--
2.39.2
Fix the following warning which appears when compiled for rv32 by using
unsigned long type instead of u64.
../drivers/firmware/efi/libstub/efi-stub-helper.c: In function 'efi_kaslr_relocate_kernel':
../drivers/firmware/efi/libstub/efi-stub-helper.c:846:28: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
846 | (u64)_end < EFI_ALLOC_LIMIT) {
Signed-off-by: Alexandre Ghiti <[email protected]>
---
drivers/firmware/efi/libstub/kaslr.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/firmware/efi/libstub/kaslr.c b/drivers/firmware/efi/libstub/kaslr.c
index be0c8ab0982a..afb857329799 100644
--- a/drivers/firmware/efi/libstub/kaslr.c
+++ b/drivers/firmware/efi/libstub/kaslr.c
@@ -130,7 +130,7 @@ efi_status_t efi_kaslr_relocate_kernel(unsigned long *image_addr,
if (!check_image_region(*image_addr, kernel_memsize)) {
efi_err("FIRMWARE BUG: Image BSS overlaps adjacent EFI memory region\n");
} else if (IS_ALIGNED(*image_addr, min_kimg_align) &&
- (u64)_end < EFI_ALLOC_LIMIT) {
+ (unsigned long)_end < EFI_ALLOC_LIMIT) {
/*
* Just execute from wherever we were loaded by the
* UEFI PE/COFF loader if the placement is suitable.
--
2.39.2
We can now use arm64 functions to handle the move of the kernel physical
mapping: if KASLR is enabled, we will try to get a random seed from the
firmware, if not possible, the kernel will be moved to a location that
suits its alignment constraints.
Signed-off-by: Alexandre Ghiti <[email protected]>
---
arch/riscv/include/asm/efi.h | 2 ++
arch/riscv/kernel/image-vars.h | 1 +
drivers/firmware/efi/libstub/riscv-stub.c | 33 +++++++++++------------
3 files changed, 18 insertions(+), 18 deletions(-)
diff --git a/arch/riscv/include/asm/efi.h b/arch/riscv/include/asm/efi.h
index 29e9a0d84b16..00b24ba55035 100644
--- a/arch/riscv/include/asm/efi.h
+++ b/arch/riscv/include/asm/efi.h
@@ -51,4 +51,6 @@ void efi_virtmap_unload(void);
unsigned long stext_offset(void);
+void efi_icache_sync(unsigned long start, unsigned long end);
+
#endif /* _ASM_EFI_H */
diff --git a/arch/riscv/kernel/image-vars.h b/arch/riscv/kernel/image-vars.h
index 15616155008c..ea1a10355ce9 100644
--- a/arch/riscv/kernel/image-vars.h
+++ b/arch/riscv/kernel/image-vars.h
@@ -27,6 +27,7 @@ __efistub__start = _start;
__efistub__start_kernel = _start_kernel;
__efistub__end = _end;
__efistub__edata = _edata;
+__efistub___init_text_end = __init_text_end;
__efistub_screen_info = screen_info;
#endif
diff --git a/drivers/firmware/efi/libstub/riscv-stub.c b/drivers/firmware/efi/libstub/riscv-stub.c
index 145c9f0ba217..c96d6dcee86c 100644
--- a/drivers/firmware/efi/libstub/riscv-stub.c
+++ b/drivers/firmware/efi/libstub/riscv-stub.c
@@ -30,32 +30,29 @@ efi_status_t handle_kernel_image(unsigned long *image_addr,
efi_loaded_image_t *image,
efi_handle_t image_handle)
{
- unsigned long kernel_size = 0;
- unsigned long preferred_addr;
+ unsigned long kernel_size, kernel_codesize, kernel_memsize;
efi_status_t status;
kernel_size = _edata - _start;
+ kernel_codesize = __init_text_end - _start;
+ kernel_memsize = kernel_size + (_end - _edata);
*image_addr = (unsigned long)_start;
- *image_size = kernel_size + (_end - _edata);
-
- /*
- * RISC-V kernel maps PAGE_OFFSET virtual address to the same physical
- * address where kernel is booted. That's why kernel should boot from
- * as low as possible to avoid wastage of memory. Currently, dram_base
- * is occupied by the firmware. So the preferred address for kernel to
- * boot is next aligned address. If preferred address is not available,
- * relocate_kernel will fall back to efi_low_alloc_above to allocate
- * lowest possible memory region as long as the address and size meets
- * the alignment constraints.
- */
- preferred_addr = EFI_KIMG_PREFERRED_ADDRESS;
- status = efi_relocate_kernel(image_addr, kernel_size, *image_size,
- preferred_addr, efi_get_kimg_min_align(),
- 0x0);
+ *image_size = kernel_memsize;
+ *reserve_size = *image_size;
+ status = efi_kaslr_relocate_kernel(image_addr,
+ reserve_addr, reserve_size,
+ kernel_size, kernel_codesize, kernel_memsize,
+ efi_kaslr_get_phys_seed(image_handle));
if (status != EFI_SUCCESS) {
efi_err("Failed to relocate kernel\n");
*image_size = 0;
}
+
return status;
}
+
+void efi_icache_sync(unsigned long start, unsigned long end)
+{
+ asm volatile ("fence.i" ::: "memory");
+}
--
2.39.2
On Fri, 21 Jul 2023 at 09:52, Alexandre Ghiti <[email protected]> wrote:
>
> Fix the following warning which appears when compiled for rv32 by using
> unsigned long type instead of u64.
>
> ../drivers/firmware/efi/libstub/efi-stub-helper.c: In function 'efi_kaslr_relocate_kernel':
> ../drivers/firmware/efi/libstub/efi-stub-helper.c:846:28: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
> 846 | (u64)_end < EFI_ALLOC_LIMIT) {
>
> Signed-off-by: Alexandre Ghiti <[email protected]>
Acked-by: Ard Biesheuvel <[email protected]>
For all I care, you can just squash this into the previous patch as well.
> ---
> drivers/firmware/efi/libstub/kaslr.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/firmware/efi/libstub/kaslr.c b/drivers/firmware/efi/libstub/kaslr.c
> index be0c8ab0982a..afb857329799 100644
> --- a/drivers/firmware/efi/libstub/kaslr.c
> +++ b/drivers/firmware/efi/libstub/kaslr.c
> @@ -130,7 +130,7 @@ efi_status_t efi_kaslr_relocate_kernel(unsigned long *image_addr,
> if (!check_image_region(*image_addr, kernel_memsize)) {
> efi_err("FIRMWARE BUG: Image BSS overlaps adjacent EFI memory region\n");
> } else if (IS_ALIGNED(*image_addr, min_kimg_align) &&
> - (u64)_end < EFI_ALLOC_LIMIT) {
> + (unsigned long)_end < EFI_ALLOC_LIMIT) {
> /*
> * Just execute from wherever we were loaded by the
> * UEFI PE/COFF loader if the placement is suitable.
> --
> 2.39.2
>