2023-02-28 15:46:40

by Alexandre Ghiti

[permalink] [raw]
Subject: [PATCH v8 0/1] riscv: Allow to downgrade paging mode from the command line

This new version gets rid of the limitation that prevented KASAN kernels
to use the newly introduced parameters.

While looking into KASLR, I fell onto commit aacd149b6238 ("arm64: head:
avoid relocating the kernel twice for KASLR"): it allows to use the fdt
functions very early in the boot process with KASAN enabled by simply
compiling a new version of those functions without instrumentation.

I had to change the handling of the command line parsing to make the
code self-contained in kernel/pi/cmd_early.c to avoid calling too many
__pi prefixed functions from outside this file.

I'll use this approach like arm64 to handle the extraction of the random
seedi from the device tree for KASLR.

base-commit: eb9be8310c58 ("RISC-V: add a spin_shadow_stack declaration")

v8:
- Fix LLVM ld warning by moving the section .init.sdata from
kernel/pi/string.c into the newly created section .init.pidata

v7:
- Rebased on top of for-next which introduces lots of errors (thanks to
the patchwork CI)
- Add __NO_FORTIFY to avoid undefined __pi_fortify_panic
- Add an alias to our newly introduced strlen
- Remove __init as sections are already prefixed in the Makefile
- Introduce new section for kernel/pi/string.c to quiet the following
warnings (once we have all the string functions, we'll be able to get
rid of this):

warning: orphan section `.init__bug_table' from `arch/riscv/kernel/pi/string.pi.o' being placed in section `.init__bug_table'
warning: orphan section `.init.srodata.cst8' from `arch/riscv/kernel/pi/string.pi.o' being placed in section `.init.srodata.cst8'

v6:
- Fix llvm warning by forward declaring set_satp_mode_from_cmdline

v5:
- Handle null command line, Thanks Björn!
- Add RB/TB from Björn

v4:
- Introduce pi/ for KASAN to work

v3:
- Massage commit log to make no4lvl clearer, as asked by Conor
- Add a note to kernel-parameters.txt regarding the impossibility to use
those parameters when KASAN is enabled, as suggested by Conor
- Add RB from Björn

v2:
- Honor CMDLINE_EXTEND and CMDLINE_FORCE as noticed by Björn


Alexandre Ghiti (1):
riscv: Allow to downgrade paging mode from the command line

.../admin-guide/kernel-parameters.txt | 5 +-
arch/riscv/kernel/Makefile | 2 +
arch/riscv/kernel/pi/Makefile | 37 +++++++++++
arch/riscv/kernel/pi/cmdline_early.c | 62 +++++++++++++++++++
arch/riscv/kernel/vmlinux.lds.S | 8 +++
arch/riscv/lib/memcpy.S | 2 +
arch/riscv/lib/memmove.S | 2 +
arch/riscv/lib/strlen.S | 1 +
arch/riscv/mm/init.c | 36 +++++++++--
9 files changed, 148 insertions(+), 7 deletions(-)
create mode 100644 arch/riscv/kernel/pi/Makefile
create mode 100644 arch/riscv/kernel/pi/cmdline_early.c

--
2.37.2



2023-02-28 15:47:42

by Alexandre Ghiti

[permalink] [raw]
Subject: [PATCH v8 1/1] riscv: Allow to downgrade paging mode from the command line

Add 2 early command line parameters that allow to downgrade satp mode
(using the same naming as x86):
- "no5lvl": use a 4-level page table (down from sv57 to sv48)
- "no4lvl": use a 3-level page table (down from sv57/sv48 to sv39)

Note that going through the device tree to get the kernel command line
works with ACPI too since the efi stub creates a device tree anyway with
the command line.

In KASAN kernels, we can't use the libfdt that early in the boot process
since we are not ready to execute instrumented functions. So instead of
using the "generic" libfdt, we compile our own versions of those functions
that are not instrumented and that are prefixed so that they do not
conflict with the generic ones. We also need the non-instrumented versions
of the string functions and the prefixed versions of memcpy/memmove.

This is largely inspired by commit aacd149b6238 ("arm64: head: avoid
relocating the kernel twice for KASLR") from which I removed compilation
flags that were not relevant to RISC-V at the moment (LTO, SCS, pie).

Signed-off-by: Alexandre Ghiti <[email protected]>
Tested-by: Björn Töpel <[email protected]>
Reviewed-by: Björn Töpel <[email protected]>
---
.../admin-guide/kernel-parameters.txt | 5 +-
arch/riscv/kernel/Makefile | 2 +
arch/riscv/kernel/pi/Makefile | 37 +++++++++++
arch/riscv/kernel/pi/cmdline_early.c | 62 +++++++++++++++++++
arch/riscv/kernel/vmlinux.lds.S | 8 +++
arch/riscv/lib/memcpy.S | 2 +
arch/riscv/lib/memmove.S | 2 +
arch/riscv/lib/strlen.S | 1 +
arch/riscv/mm/init.c | 36 +++++++++--
9 files changed, 148 insertions(+), 7 deletions(-)
create mode 100644 arch/riscv/kernel/pi/Makefile
create mode 100644 arch/riscv/kernel/pi/cmdline_early.c

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 6cfa6e3996cf..1d8fc66d2b0d 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -3578,7 +3578,10 @@
emulation library even if a 387 maths coprocessor
is present.

- no5lvl [X86-64] Disable 5-level paging mode. Forces
+ no4lvl [RISCV] Disable 4-level and 5-level paging modes. Forces
+ kernel to use 3-level paging instead.
+
+ no5lvl [X86-64,RISCV] Disable 5-level paging mode. Forces
kernel to use 4-level paging instead.

nofsgsbase [X86] Disables FSGSBASE instructions.
diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
index 4cf303a779ab..aa22f87faeae 100644
--- a/arch/riscv/kernel/Makefile
+++ b/arch/riscv/kernel/Makefile
@@ -89,3 +89,5 @@ obj-$(CONFIG_EFI) += efi.o
obj-$(CONFIG_COMPAT) += compat_syscall_table.o
obj-$(CONFIG_COMPAT) += compat_signal.o
obj-$(CONFIG_COMPAT) += compat_vdso/
+
+obj-$(CONFIG_64BIT) += pi/
diff --git a/arch/riscv/kernel/pi/Makefile b/arch/riscv/kernel/pi/Makefile
new file mode 100644
index 000000000000..4002ed94b6d3
--- /dev/null
+++ b/arch/riscv/kernel/pi/Makefile
@@ -0,0 +1,37 @@
+# SPDX-License-Identifier: GPL-2.0
+# This file was copied from arm64/kernel/pi/Makefile.
+
+KBUILD_CFLAGS := $(subst $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) \
+ -Os -DDISABLE_BRANCH_PROFILING $(DISABLE_STACKLEAK_PLUGIN) \
+ $(call cc-option,-mbranch-protection=none) \
+ -I$(srctree)/scripts/dtc/libfdt -fno-stack-protector \
+ -D__DISABLE_EXPORTS -ffreestanding \
+ -fno-asynchronous-unwind-tables -fno-unwind-tables \
+ $(call cc-option,-fno-addrsig)
+
+CFLAGS_cmdline_early.o += -D__NO_FORTIFY
+CFLAGS_lib-fdt_ro.o += -D__NO_FORTIFY
+
+GCOV_PROFILE := n
+KASAN_SANITIZE := n
+KCSAN_SANITIZE := n
+UBSAN_SANITIZE := n
+KCOV_INSTRUMENT := n
+
+$(obj)/%.pi.o: OBJCOPYFLAGS := --prefix-symbols=__pi_ \
+ --remove-section=.note.gnu.property \
+ --prefix-alloc-sections=.init
+$(obj)/%.pi.o: $(obj)/%.o FORCE
+ $(call if_changed,objcopy)
+
+$(obj)/lib-%.o: $(srctree)/lib/%.c FORCE
+ $(call if_changed_rule,cc_o_c)
+
+$(obj)/string.o: $(srctree)/lib/string.c FORCE
+ $(call if_changed_rule,cc_o_c)
+
+$(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
+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
new file mode 100644
index 000000000000..05652d13c746
--- /dev/null
+++ b/arch/riscv/kernel/pi/cmdline_early.c
@@ -0,0 +1,62 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include <linux/types.h>
+#include <linux/init.h>
+#include <linux/libfdt.h>
+#include <linux/string.h>
+#include <asm/pgtable.h>
+#include <asm/setup.h>
+
+static char early_cmdline[COMMAND_LINE_SIZE];
+
+/*
+ * 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 set_satp_mode_from_cmdline(uintptr_t dtb_pa);
+
+static char *get_early_cmdline(uintptr_t dtb_pa)
+{
+ const char *fdt_cmdline = NULL;
+ unsigned int fdt_cmdline_size = 0;
+ int chosen_node;
+
+ if (!IS_ENABLED(CONFIG_CMDLINE_FORCE)) {
+ chosen_node = fdt_path_offset((void *)dtb_pa, "/chosen");
+ if (chosen_node >= 0) {
+ fdt_cmdline = fdt_getprop((void *)dtb_pa, chosen_node,
+ "bootargs", NULL);
+ if (fdt_cmdline) {
+ fdt_cmdline_size = strlen(fdt_cmdline);
+ strscpy(early_cmdline, fdt_cmdline,
+ COMMAND_LINE_SIZE);
+ }
+ }
+ }
+
+ if (IS_ENABLED(CONFIG_CMDLINE_EXTEND) ||
+ IS_ENABLED(CONFIG_CMDLINE_FORCE) ||
+ fdt_cmdline_size == 0 /* CONFIG_CMDLINE_FALLBACK */) {
+ strncat(early_cmdline, CONFIG_CMDLINE,
+ COMMAND_LINE_SIZE - fdt_cmdline_size);
+ }
+
+ return early_cmdline;
+}
+
+static u64 match_noXlvl(char *cmdline)
+{
+ if (strstr(cmdline, "no4lvl"))
+ return SATP_MODE_48;
+ else if (strstr(cmdline, "no5lvl"))
+ return SATP_MODE_57;
+
+ return 0;
+}
+
+u64 set_satp_mode_from_cmdline(uintptr_t dtb_pa)
+{
+ char *cmdline = get_early_cmdline(dtb_pa);
+
+ return match_noXlvl(cmdline);
+}
diff --git a/arch/riscv/kernel/vmlinux.lds.S b/arch/riscv/kernel/vmlinux.lds.S
index b2813b5e47a0..21d402e65e26 100644
--- a/arch/riscv/kernel/vmlinux.lds.S
+++ b/arch/riscv/kernel/vmlinux.lds.S
@@ -87,6 +87,14 @@ SECTIONS
/* Start of init data section */
__init_data_begin = .;
INIT_DATA_SECTION(16)
+
+ /* Those sections result from the compilation of kernel/pi/string.c */
+ .init.pidata : {
+ *(.init.srodata.cst8*)
+ *(.init__bug_table*)
+ *(.init.sdata*)
+ }
+
.init.bss : {
*(.init.bss) /* from the EFI stub */
}
diff --git a/arch/riscv/lib/memcpy.S b/arch/riscv/lib/memcpy.S
index 51ab716253fa..1a40d01a9543 100644
--- a/arch/riscv/lib/memcpy.S
+++ b/arch/riscv/lib/memcpy.S
@@ -106,3 +106,5 @@ WEAK(memcpy)
6:
ret
END(__memcpy)
+SYM_FUNC_ALIAS(__pi_memcpy, __memcpy)
+SYM_FUNC_ALIAS(__pi___memcpy, __memcpy)
diff --git a/arch/riscv/lib/memmove.S b/arch/riscv/lib/memmove.S
index e0609e1f0864..838ff2022fe3 100644
--- a/arch/riscv/lib/memmove.S
+++ b/arch/riscv/lib/memmove.S
@@ -314,3 +314,5 @@ return_from_memmove:

SYM_FUNC_END(memmove)
SYM_FUNC_END(__memmove)
+SYM_FUNC_ALIAS(__pi_memmove, __memmove)
+SYM_FUNC_ALIAS(__pi___memmove, __memmove)
diff --git a/arch/riscv/lib/strlen.S b/arch/riscv/lib/strlen.S
index 8345ceeee3f6..ed43be976134 100644
--- a/arch/riscv/lib/strlen.S
+++ b/arch/riscv/lib/strlen.S
@@ -131,3 +131,4 @@ strlen_zbb:
.option pop
#endif
SYM_FUNC_END(strlen)
+SYM_FUNC_ALIAS(__pi_strlen, strlen)
diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c
index 478d6763a01a..70c7fc566f63 100644
--- a/arch/riscv/mm/init.c
+++ b/arch/riscv/mm/init.c
@@ -732,6 +732,8 @@ static __init pgprot_t pgprot_from_va(uintptr_t va)
#endif /* CONFIG_STRICT_KERNEL_RWX */

#if defined(CONFIG_64BIT) && !defined(CONFIG_XIP_KERNEL)
+u64 __pi_set_satp_mode_from_cmdline(uintptr_t dtb_pa);
+
static void __init disable_pgtable_l5(void)
{
pgtable_l5_enabled = false;
@@ -746,17 +748,39 @@ static void __init disable_pgtable_l4(void)
satp_mode = SATP_MODE_39;
}

+static int __init print_no4lvl(char *p)
+{
+ pr_info("Disabled 4-level and 5-level paging");
+ return 0;
+}
+early_param("no4lvl", print_no4lvl);
+
+static int __init print_no5lvl(char *p)
+{
+ pr_info("Disabled 5-level paging");
+ return 0;
+}
+early_param("no5lvl", print_no5lvl);
+
/*
* There is a simple way to determine if 4-level is supported by the
* underlying hardware: establish 1:1 mapping in 4-level page table mode
* then read SATP to see if the configuration was taken into account
* meaning sv48 is supported.
*/
-static __init void set_satp_mode(void)
+static __init void set_satp_mode(uintptr_t dtb_pa)
{
u64 identity_satp, hw_satp;
uintptr_t set_satp_mode_pmd = ((unsigned long)set_satp_mode) & PMD_MASK;
- bool check_l4 = false;
+ u64 satp_mode_cmdline = __pi_set_satp_mode_from_cmdline(dtb_pa);
+
+ if (satp_mode_cmdline == SATP_MODE_57) {
+ disable_pgtable_l5();
+ } else if (satp_mode_cmdline == SATP_MODE_48) {
+ disable_pgtable_l5();
+ disable_pgtable_l4();
+ return;
+ }

create_p4d_mapping(early_p4d,
set_satp_mode_pmd, (uintptr_t)early_pud,
@@ -775,7 +799,8 @@ static __init void set_satp_mode(void)
retry:
create_pgd_mapping(early_pg_dir,
set_satp_mode_pmd,
- check_l4 ? (uintptr_t)early_pud : (uintptr_t)early_p4d,
+ pgtable_l5_enabled ?
+ (uintptr_t)early_p4d : (uintptr_t)early_pud,
PGDIR_SIZE, PAGE_TABLE);

identity_satp = PFN_DOWN((uintptr_t)&early_pg_dir) | satp_mode;
@@ -786,9 +811,8 @@ static __init void set_satp_mode(void)
local_flush_tlb_all();

if (hw_satp != identity_satp) {
- if (!check_l4) {
+ if (pgtable_l5_enabled) {
disable_pgtable_l5();
- check_l4 = true;
memset(early_pg_dir, 0, PAGE_SIZE);
goto retry;
}
@@ -979,7 +1003,7 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa)
#endif

#if defined(CONFIG_64BIT) && !defined(CONFIG_XIP_KERNEL)
- set_satp_mode();
+ set_satp_mode(dtb_pa);
#endif

kernel_map.va_pa_offset = PAGE_OFFSET - kernel_map.phys_addr;
--
2.37.2


2023-03-29 18:03:47

by Palmer Dabbelt

[permalink] [raw]
Subject: Re: [PATCH v8 0/1] riscv: Allow to downgrade paging mode from the command line

On Tue, 28 Feb 2023 07:46:28 PST (-0800), [email protected] wrote:
> This new version gets rid of the limitation that prevented KASAN kernels
> to use the newly introduced parameters.
>
> While looking into KASLR, I fell onto commit aacd149b6238 ("arm64: head:
> avoid relocating the kernel twice for KASLR"): it allows to use the fdt
> functions very early in the boot process with KASAN enabled by simply
> compiling a new version of those functions without instrumentation.
>
> I had to change the handling of the command line parsing to make the
> code self-contained in kernel/pi/cmd_early.c to avoid calling too many
> __pi prefixed functions from outside this file.
>
> I'll use this approach like arm64 to handle the extraction of the random
> seedi from the device tree for KASLR.
>
> base-commit: eb9be8310c58 ("RISC-V: add a spin_shadow_stack declaration")

I have that in my tree, I'm still getting a bunch of boot failures
though. I can try to get a list if you're not seeing any, but I'm kind
of stuck on some other stuff for a bit...

> v8:
> - Fix LLVM ld warning by moving the section .init.sdata from
> kernel/pi/string.c into the newly created section .init.pidata
>
> v7:
> - Rebased on top of for-next which introduces lots of errors (thanks to
> the patchwork CI)
> - Add __NO_FORTIFY to avoid undefined __pi_fortify_panic
> - Add an alias to our newly introduced strlen
> - Remove __init as sections are already prefixed in the Makefile
> - Introduce new section for kernel/pi/string.c to quiet the following
> warnings (once we have all the string functions, we'll be able to get
> rid of this):
>
> warning: orphan section `.init__bug_table' from `arch/riscv/kernel/pi/string.pi.o' being placed in section `.init__bug_table'
> warning: orphan section `.init.srodata.cst8' from `arch/riscv/kernel/pi/string.pi.o' being placed in section `.init.srodata.cst8'
>
> v6:
> - Fix llvm warning by forward declaring set_satp_mode_from_cmdline
>
> v5:
> - Handle null command line, Thanks Björn!
> - Add RB/TB from Björn
>
> v4:
> - Introduce pi/ for KASAN to work
>
> v3:
> - Massage commit log to make no4lvl clearer, as asked by Conor
> - Add a note to kernel-parameters.txt regarding the impossibility to use
> those parameters when KASAN is enabled, as suggested by Conor
> - Add RB from Björn
>
> v2:
> - Honor CMDLINE_EXTEND and CMDLINE_FORCE as noticed by Björn
>
>
> Alexandre Ghiti (1):
> riscv: Allow to downgrade paging mode from the command line
>
> .../admin-guide/kernel-parameters.txt | 5 +-
> arch/riscv/kernel/Makefile | 2 +
> arch/riscv/kernel/pi/Makefile | 37 +++++++++++
> arch/riscv/kernel/pi/cmdline_early.c | 62 +++++++++++++++++++
> arch/riscv/kernel/vmlinux.lds.S | 8 +++
> arch/riscv/lib/memcpy.S | 2 +
> arch/riscv/lib/memmove.S | 2 +
> arch/riscv/lib/strlen.S | 1 +
> arch/riscv/mm/init.c | 36 +++++++++--
> 9 files changed, 148 insertions(+), 7 deletions(-)
> create mode 100644 arch/riscv/kernel/pi/Makefile
> create mode 100644 arch/riscv/kernel/pi/cmdline_early.c

2023-03-30 04:18:45

by Alexandre Ghiti

[permalink] [raw]
Subject: Re: [PATCH v8 0/1] riscv: Allow to downgrade paging mode from the command line

On 3/29/23 19:59, Palmer Dabbelt wrote:
> On Tue, 28 Feb 2023 07:46:28 PST (-0800), [email protected] wrote:
>> This new version gets rid of the limitation that prevented KASAN kernels
>> to use the newly introduced parameters.
>>
>> While looking into KASLR, I fell onto commit aacd149b6238 ("arm64: head:
>> avoid relocating the kernel twice for KASLR"): it allows to use the fdt
>> functions very early in the boot process with KASAN enabled by simply
>> compiling a new version of those functions without instrumentation.
>>
>> I had to change the handling of the command line parsing to make the
>> code self-contained in kernel/pi/cmd_early.c to avoid calling too many
>> __pi prefixed functions from outside this file.
>>
>> I'll use this approach like arm64 to handle the extraction of the random
>> seedi from the device tree for KASLR.
>>
>> base-commit: eb9be8310c58 ("RISC-V: add a spin_shadow_stack
>> declaration")
>
> I have that in my tree, I'm still getting a bunch of boot failures
> though.  I can try to get a list if you're not seeing any, but I'm
> kind of stuck on some other stuff for a bit...


I sent yesterday a v9 which fixes the medlow boot issues you mentioned,
I did not have any boot failures after that on my side.


>
>> v8:
>> - Fix LLVM ld warning by moving the section .init.sdata from
>>   kernel/pi/string.c into the newly created section .init.pidata
>>
>> v7:
>> - Rebased on top of for-next which introduces lots of errors (thanks to
>>   the patchwork CI)
>> - Add __NO_FORTIFY to avoid undefined __pi_fortify_panic
>> - Add an alias to our newly introduced strlen
>> - Remove __init as sections are already prefixed in the Makefile
>> - Introduce new section for kernel/pi/string.c to quiet the following
>>   warnings (once we have all the string functions, we'll be able to get
>>   rid of this):
>>
>> warning: orphan section `.init__bug_table' from
>> `arch/riscv/kernel/pi/string.pi.o' being placed in section
>> `.init__bug_table'
>> warning: orphan section `.init.srodata.cst8' from
>> `arch/riscv/kernel/pi/string.pi.o' being placed in section
>> `.init.srodata.cst8'
>>
>> v6:
>> - Fix llvm warning by forward declaring set_satp_mode_from_cmdline
>>
>> v5:
>> - Handle null command line, Thanks Björn!
>> - Add RB/TB from Björn
>>
>> v4:
>> - Introduce pi/ for KASAN to work
>>
>> v3:
>> - Massage commit log to make no4lvl clearer, as asked by Conor
>> - Add a note to kernel-parameters.txt regarding the impossibility to use
>>   those parameters when KASAN is enabled, as suggested by Conor
>> - Add RB from Björn
>>
>> v2:
>> - Honor CMDLINE_EXTEND and CMDLINE_FORCE as noticed by Björn
>>
>>
>> Alexandre Ghiti (1):
>>   riscv: Allow to downgrade paging mode from the command line
>>
>>  .../admin-guide/kernel-parameters.txt         |  5 +-
>>  arch/riscv/kernel/Makefile                    |  2 +
>>  arch/riscv/kernel/pi/Makefile                 | 37 +++++++++++
>>  arch/riscv/kernel/pi/cmdline_early.c          | 62 +++++++++++++++++++
>>  arch/riscv/kernel/vmlinux.lds.S               |  8 +++
>>  arch/riscv/lib/memcpy.S                       |  2 +
>>  arch/riscv/lib/memmove.S                      |  2 +
>>  arch/riscv/lib/strlen.S                       |  1 +
>>  arch/riscv/mm/init.c                          | 36 +++++++++--
>>  9 files changed, 148 insertions(+), 7 deletions(-)
>>  create mode 100644 arch/riscv/kernel/pi/Makefile
>>  create mode 100644 arch/riscv/kernel/pi/cmdline_early.c
>
> _______________________________________________
> linux-riscv mailing list
> [email protected]
> http://lists.infradead.org/mailman/listinfo/linux-riscv