Basics and overview ===================
Software with larger attack surfaces (e.g. network facing apps like databases, browsers or apps relying on browser runtimes) suffer from memory corruption issues which can be utilized by attackers to bend control flow of the program to eventually gain control (by making their payload executable). Attackers are able to perform such attacks by leveraging call-sites which rely on indirect calls or return sites which rely on obtaining return address from stack memory.
To mitigate such attacks, risc-v extension zicfilp enforces that all indirect calls must land on a landing pad instruction `lpad` else cpu will raise software check exception (a new cpu exception cause code on riscv). Similarly for return flow, risc-v extension zicfiss extends architecture with
- `sspush` instruction to push return address on a shadow stack - `sspopchk` instruction to pop return address from shadow stack and compare with input operand (i.e. return address on stack) - `sspopchk` to raise software check exception if comparision above was a mismatch - Protection mechanism using which shadow stack is not writeable via regular store instructions
More information an details can be found at extensions github repo [1].
Equivalent to landing pad (zicfilp) on x86 is `ENDBRANCH` instruction in Intel CET [3] and branch target identification (BTI) [4] on arm. Similarly x86's Intel CET has shadow stack [5] and arm64 has guarded control stack (GCS) [6] which are very similar to risc-v's zicfiss shadow stack.
x86 and arm64 support for user mode shadow stack is already in mainline.
Kernel awareness for user control flow integrity ================================================
This series picks up Samuel Holland's envcfg changes [2] as well. So if those are being applied independently, they should be removed from this series.
Enabling:
In order to maintain compatibility and not break anything in user mode, kernel doesn't enable control flow integrity cpu extensions on binary by default. Instead exposes a prctl interface to enable, disable and lock the shadow stack or landing pad feature for a task. This allows userspace (loader) to enumerate if all objects in its address space are compiled with shadow stack and landing pad support and accordingly enable the feature. Additionally if a subsequent `dlopen` happens on a library, user mode can take a decision again to disable the feature (if incoming library is not compiled with support) OR terminate the task (if user mode policy is strict to have all objects in address space to be compiled with control flow integirty cpu feature). prctl to enable shadow stack results in allocating shadow stack from virtual memory and activating for user address space. x86 and arm64 are also following same direction due to similar reason(s).
clone/fork:
On clone and fork, cfi state for task is inherited by child. Shadow stack is part of virtual memory and is a writeable memory from kernel perspective (writeable via a restricted set of instructions aka shadow stack instructions) Thus kernel changes ensure that this memory is converted into read-only when fork/clone happens and COWed when fault is taken due to sspush, sspopchk or ssamoswap. In case `CLONE_VM` is specified and shadow stack is to be enabled, kernel will automatically allocate a shadow stack for that clone call.
map_shadow_stack:
x86 introduced `map_shadow_stack` system call to allow user space to explicitly map shadow stack memory in its address space. It is useful to allocate shadow for different contexts managed by a single thread (green threads or contexts) risc-v implements this system call as well.
signal management:
If shadow stack is enabled for a task, kernel performs an asynchronous control flow diversion to deliver the signal and eventually expects userspace to issue sigreturn so that original execution can be resumed. Even though resume context is prepared by kernel, it is in user space memory and is subject to memory corruption and corruption bugs can be utilized by attacker in this race window to perform arbitrary sigreturn and eventually bypass cfi mechanism. Another issue is how to ensure that cfi related state on sigcontext area is not trampled by legacy apps or apps compiled with old kernel headers.
In order to mitigate control-flow hijacting, kernel prepares a token and place it on shadow stack before signal delivery and places address of token in sigcontext structure. During sigreturn, kernel obtains address of token from sigcontext struture, reads token from shadow stack and validates it and only then allow sigreturn to succeed. Compatiblity issue is solved by adopting dynamic sigcontext management introduced for vector extension. This series re-factor the code little bit to allow future sigcontext management easy (as proposed by Andy Chiu from SiFive)
config and compilation:
Introduce a new risc-v config option `CONFIG_RISCV_USER_CFI`. Selecting this config option picks the kernel support for user control flow integrity. This optin is presented only if toolchain has shadow stack and landing pad support. And is on purpose guarded by toolchain support. Reason being that eventually vDSO also needs to be compiled in with shadow stack and landing pad support. vDSO compile patches are not included as of now because landing pad labeling scheme is yet to settle for usermode runtime.
To get more information on kernel interactions with respect to zicfilp and zicfiss, patch series adds documentation for `zicfilp` and `zicfiss` in following: Documentation/arch/riscv/zicfiss.rst Documentation/arch/riscv/zicfilp.rst
How to test this series =======================
Toolchain --------- $ git clone git@github.com:sifive/riscv-gnu-toolchain.git -b cfi-dev $ riscv-gnu-toolchain/configure --prefix=<path-to-where-to-build> --with-arch=rv64gc_zicfilp_zicfiss --enable-linux --disable-gdb --with-extra-multilib-test="rv64gc_zicfilp_zicfiss-lp64d:-static" $ make -j$(nproc)
Qemu ---- Get the lastest qemu $ cd qemu $ mkdir build $ cd build $ ../configure --target-list=riscv64-softmmu $ make -j$(nproc)
Opensbi ------- $ git clone git@github.com:deepak0414/opensbi.git -b v6_cfi_spec_split_opensbi $ make CROSS_COMPILE=<your riscv toolchain> -j$(nproc) PLATFORM=generic
Linux ----- Running defconfig is fine. CFI is enabled by default if the toolchain supports it.
$ make ARCH=riscv CROSS_COMPILE=<path-to-cfi-riscv-gnu-toolchain>/build/bin/riscv64-unknown-linux-gnu- -j$(nproc) defconfig $ make ARCH=riscv CROSS_COMPILE=<path-to-cfi-riscv-gnu-toolchain>/build/bin/riscv64-unknown-linux-gnu- -j$(nproc)
In case you're building your own rootfs using toolchain, please make sure you pick following patch to ensure that vDSO compiled with lpad and shadow stack.
"arch/riscv: compile vdso with landing pad"
Branch where above patch can be picked https://github.com/deepak0414/linux-riscv-cfi/tree/vdso_user_cfi_v6.12-rc1
Running -------
Modify your qemu command to have: -bios <path-to-cfi-opensbi>/build/platform/generic/firmware/fw_dynamic.bin -cpu rv64,zicfilp=true,zicfiss=true,zimop=true,zcmop=true
vDSO related Opens (in the flux) =================================
I am listing these opens for laying out plan and what to expect in future patch sets. And of course for the sake of discussion.
Shadow stack and landing pad enabling in vDSO ---------------------------------------------- vDSO must have shadow stack and landing pad support compiled in for task to have shadow stack and landing pad support. This patch series doesn't enable that (yet). Enabling shadow stack support in vDSO should be straight forward (intend to do that in next versions of patch set). Enabling landing pad support in vDSO requires some collaboration with toolchain folks to follow a single label scheme for all object binaries. This is necessary to ensure that all indirect call-sites are setting correct label and target landing pads are decorated with same label scheme.
How many vDSOs --------------- Shadow stack instructions are carved out of zimop (may be operations) and if CPU doesn't implement zimop, they're illegal instructions. Kernel could be running on a CPU which may or may not implement zimop. And thus kernel will have to carry 2 different vDSOs and expose the appropriate one depending on whether CPU implements zimop or not.
References ========== [1] - https://github.com/riscv/riscv-cfi [2] - https://lore.kernel.org/all/20240814081126.956287-1-samuel.holland@sifive.co... [3] - https://lwn.net/Articles/889475/ [4] - https://developer.arm.com/documentation/109576/0100/Branch-Target-Identifica... [5] - https://www.intel.com/content/dam/develop/external/us/en/documents/catc17-in... [6] - https://lwn.net/Articles/940403/
--- changelog --------- v12: - It seems like I had accidently squashed arch agnostic indirect branch tracking prctl and riscv implementation of those prctls. Split them again. - set_shstk_status/set_indir_lp_status perform CSR writes only when CPU support is available. As suggested by Zong Li. - Some minor clean up in kselftests as suggested by Zong Li.
v11: - patch "arch/riscv: compile vdso with landing pad" was unconditionally selecting `_zicfilp` for vDSO compile. fixed that. Changed `lpad 1` to to `lpad 0`. v10: - dropped "mm: helper `is_shadow_stack_vma` to check shadow stack vma". This patch is not that interesting to this patch series for risc-v. There are instances in arch directories where VM_SHADOW_STACK flag is anyways used. Dropping this patch to expedite merging in riscv tree. - Took suggestions from `Clement` on "riscv: zicfiss / zicfilp enumeration" to validate presence of cfi based on config. - Added a patch for vDSO to have `lpad 0`. I had omitted this earlier to make sure we add single vdso object with cfi enabled. But a vdso object with scheme of zero labeled landing pad is least common denominator and should work with all objects of zero labeled as well as function-signature labeled objects.
v9: - rebased on master (39a803b754d5 fix braino in "9p: fix ->rename_sem exclusion") - dropped "mm: Introduce ARCH_HAS_USER_SHADOW_STACK" (master has it from arm64/gcs) - dropped "prctl: arch-agnostic prctl for shadow stack" (master has it from arm64/gcs)
v8: - rebased on palmer/for-next - dropped samuel holland's `envcfg` context switch patches. they are in parlmer/for-next
v7: - Removed "riscv/Kconfig: enable HAVE_EXIT_THREAD for riscv" Instead using `deactivate_mm` flow to clean up. see here for more context https://lore.kernel.org/all/20230908203655.543765-1-rick.p.edgecombe@intel.c... - Changed the header include in `kselftest`. Hopefully this fixes compile issue faced by Zong Li at SiFive. - Cleaned up an orphaned change to `mm/mmap.c` in below patch "riscv/mm : ensure PROT_WRITE leads to VM_READ | VM_WRITE" - Lock interfaces for shadow stack and indirect branch tracking expect arg == 0 Any future evolution of this interface should accordingly define how arg should be setup. - `mm/map.c` has an instance of using `VM_SHADOW_STACK`. Fixed it to use helper `is_shadow_stack_vma`. - Link to v6: https://lore.kernel.org/r/20241008-v5_user_cfi_series-v6-0-60d9fe073f37@rivo...
v6: - Picked up Samuel Holland's changes as is with `envcfg` placed in `thread` instead of `thread_info` - fixed unaligned newline escapes in kselftest - cleaned up messages in kselftest and included test output in commit message - fixed a bug in clone path reported by Zong Li - fixed a build issue if CONFIG_RISCV_ISA_V is not selected (this was introduced due to re-factoring signal context management code)
v5: - rebased on v6.12-rc1 - Fixed schema related issues in device tree file - Fixed some of the documentation related issues in zicfilp/ss.rst (style issues and added index) - added `SHADOW_STACK_SET_MARKER` so that implementation can define base of shadow stack. - Fixed warnings on definitions added in usercfi.h when CONFIG_RISCV_USER_CFI is not selected. - Adopted context header based signal handling as proposed by Andy Chiu - Added support for enabling kernel mode access to shadow stack using FWFT (https://github.com/riscv-non-isa/riscv-sbi-doc/blob/master/src/ext-firmware-...) - Link to v5: https://lore.kernel.org/r/20241001-v5_user_cfi_series-v1-0-3ba65b6e550f@rivo... (Note: I had an issue in my workflow due to which version number wasn't picked up correctly while sending out patches)
v4: - rebased on 6.11-rc6 - envcfg: Converged with Samuel Holland's patches for envcfg management on per- thread basis. - vma_is_shadow_stack is renamed to is_vma_shadow_stack - picked up Mark Brown's `ARCH_HAS_USER_SHADOW_STACK` patch - signal context: using extended context management to maintain compatibility. - fixed `-Wmissing-prototypes` compiler warnings for prctl functions - Documentation fixes and amending typos. - Link to v4: https://lore.kernel.org/all/20240912231650.3740732-1-debug@rivosinc.com/
v3: - envcfg logic to pick up base envcfg had a bug where `ENVCFG_CBZE` could have been picked on per task basis, even though CPU didn't implement it. Fixed in this series.
- dt-bindings As suggested, split into separate commit. fixed the messaging that spec is in public review
- arch_is_shadow_stack change arch_is_shadow_stack changed to vma_is_shadow_stack
- hwprobe zicfiss / zicfilp if present will get enumerated in hwprobe
- selftests As suggested, added object and binary filenames to .gitignore Selftest binary anyways need to be compiled with cfi enabled compiler which will make sure that landing pad and shadow stack are enabled. Thus removed separate enable/disable tests. Cleaned up tests a bit.
- Link to v3: https://lore.kernel.org/lkml/20240403234054.2020347-1-debug@rivosinc.com/
v2: - Using config `CONFIG_RISCV_USER_CFI`, kernel support for riscv control flow integrity for user mode programs can be compiled in the kernel.
- Enabling of control flow integrity for user programs is left to user runtime
- This patch series introduces arch agnostic `prctls` to enable shadow stack and indirect branch tracking. And implements them on riscv.
---
--- Changes in v12: - EDITME: describe what is new in this series revision. - EDITME: use bulletpoints and terse descriptions. - Link to v11: https://lore.kernel.org/r/20250310-v5_user_cfi_series-v11-0-86b36cbfb910@riv...
Changes in v11: - EDITME: describe what is new in this series revision. - EDITME: use bulletpoints and terse descriptions. - Link to v10: https://lore.kernel.org/r/20250210-v5_user_cfi_series-v10-0-163dcfa31c60@riv...
--- Andy Chiu (1): riscv: signal: abstract header saving for setup_sigcontext
Clément Léger (1): riscv: Add Firmware Feature SBI extensions definitions
Deepak Gupta (25): mm: VM_SHADOW_STACK definition for riscv dt-bindings: riscv: zicfilp and zicfiss in dt-bindings (extensions.yaml) riscv: zicfiss / zicfilp enumeration riscv: zicfiss / zicfilp extension csr and bit definitions riscv: usercfi state for task and save/restore of CSR_SSP on trap entry/exit riscv/mm : ensure PROT_WRITE leads to VM_READ | VM_WRITE riscv mm: manufacture shadow stack pte riscv mmu: teach pte_mkwrite to manufacture shadow stack PTEs riscv mmu: write protect and shadow stack riscv/mm: Implement map_shadow_stack() syscall riscv/shstk: If needed allocate a new shadow stack on clone riscv: Implements arch agnostic shadow stack prctls prctl: arch-agnostic prctl for indirect branch tracking riscv: Implements arch agnostic indirect branch tracking prctls riscv/traps: Introduce software check exception riscv/signal: save and restore of shadow stack for signal riscv/kernel: update __show_regs to print shadow stack register riscv/ptrace: riscv cfi status and state via ptrace and in core files riscv/hwprobe: zicfilp / zicfiss enumeration in hwprobe riscv: enable kernel access to shadow stack memory via FWFT sbi call riscv: kernel command line option to opt out of user cfi riscv: create a config for shadow stack and landing pad instr support riscv: Documentation for landing pad / indirect branch tracking riscv: Documentation for shadow stack on riscv kselftest/riscv: kselftest for user mode cfi
Jim Shu (1): arch/riscv: compile vdso with landing pad
Documentation/arch/riscv/index.rst | 2 + Documentation/arch/riscv/zicfilp.rst | 115 +++++ Documentation/arch/riscv/zicfiss.rst | 176 +++++++ .../devicetree/bindings/riscv/extensions.yaml | 14 + arch/riscv/Kconfig | 20 + arch/riscv/Makefile | 5 +- arch/riscv/include/asm/asm-prototypes.h | 1 + arch/riscv/include/asm/assembler.h | 44 ++ arch/riscv/include/asm/cpufeature.h | 13 + arch/riscv/include/asm/csr.h | 16 + arch/riscv/include/asm/entry-common.h | 2 + arch/riscv/include/asm/hwcap.h | 2 + arch/riscv/include/asm/mman.h | 25 + arch/riscv/include/asm/mmu_context.h | 7 + arch/riscv/include/asm/pgtable.h | 30 +- arch/riscv/include/asm/processor.h | 2 + arch/riscv/include/asm/sbi.h | 26 + arch/riscv/include/asm/thread_info.h | 3 + arch/riscv/include/asm/usercfi.h | 89 ++++ arch/riscv/include/asm/vector.h | 3 + arch/riscv/include/uapi/asm/hwprobe.h | 2 + arch/riscv/include/uapi/asm/ptrace.h | 22 + arch/riscv/include/uapi/asm/sigcontext.h | 1 + arch/riscv/kernel/Makefile | 1 + arch/riscv/kernel/asm-offsets.c | 8 + arch/riscv/kernel/cpufeature.c | 13 + arch/riscv/kernel/entry.S | 31 +- arch/riscv/kernel/head.S | 12 + arch/riscv/kernel/process.c | 26 +- arch/riscv/kernel/ptrace.c | 83 ++++ arch/riscv/kernel/signal.c | 142 +++++- arch/riscv/kernel/sys_hwprobe.c | 2 + arch/riscv/kernel/sys_riscv.c | 10 + arch/riscv/kernel/traps.c | 43 ++ arch/riscv/kernel/usercfi.c | 530 +++++++++++++++++++++ arch/riscv/kernel/vdso/Makefile | 12 + arch/riscv/kernel/vdso/flush_icache.S | 4 + arch/riscv/kernel/vdso/getcpu.S | 4 + arch/riscv/kernel/vdso/rt_sigreturn.S | 4 + arch/riscv/kernel/vdso/sys_hwprobe.S | 4 + arch/riscv/mm/init.c | 2 +- arch/riscv/mm/pgtable.c | 17 + include/linux/cpu.h | 4 + include/linux/mm.h | 7 + include/uapi/linux/elf.h | 1 + include/uapi/linux/prctl.h | 27 ++ kernel/sys.c | 30 ++ tools/testing/selftests/riscv/Makefile | 2 +- tools/testing/selftests/riscv/cfi/.gitignore | 3 + tools/testing/selftests/riscv/cfi/Makefile | 10 + tools/testing/selftests/riscv/cfi/cfi_rv_test.h | 82 ++++ tools/testing/selftests/riscv/cfi/riscv_cfi_test.c | 78 +++ tools/testing/selftests/riscv/cfi/shadowstack.c | 385 +++++++++++++++ tools/testing/selftests/riscv/cfi/shadowstack.h | 27 ++ 54 files changed, 2195 insertions(+), 29 deletions(-) --- base-commit: 39a803b754d5224a3522016b564113ee1e4091b2 change-id: 20240930-v5_user_cfi_series-3dc332f8f5b2 -- - debug
VM_HIGH_ARCH_5 is used for riscv
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- include/linux/mm.h | 7 +++++++ 1 file changed, 7 insertions(+)
diff --git a/include/linux/mm.h b/include/linux/mm.h index 7b1068ddcbb7..1ef231cbc8fe 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -378,6 +378,13 @@ extern unsigned int kobjsize(const void *objp); # define VM_SHADOW_STACK VM_HIGH_ARCH_6 #endif
+#if defined(CONFIG_RISCV_USER_CFI) +/* + * Following x86 and picking up the same bitpos. + */ +# define VM_SHADOW_STACK VM_HIGH_ARCH_5 +#endif + #ifndef VM_SHADOW_STACK # define VM_SHADOW_STACK VM_NONE #endif
Hi Deepak,
On 14/03/2025 22:39, Deepak Gupta wrote:
Reviewed-by: Alexandre Ghiti alexghiti@rivosinc.com
Thanks,
Alex
Make an entry for cfi extensions in extensions.yaml.
Signed-off-by: Deepak Gupta debug@rivosinc.com Acked-by: Rob Herring (Arm) robh@kernel.org --- Documentation/devicetree/bindings/riscv/extensions.yaml | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
diff --git a/Documentation/devicetree/bindings/riscv/extensions.yaml b/Documentation/devicetree/bindings/riscv/extensions.yaml index a63b994e0763..9b9024dbc8d2 100644 --- a/Documentation/devicetree/bindings/riscv/extensions.yaml +++ b/Documentation/devicetree/bindings/riscv/extensions.yaml @@ -426,6 +426,20 @@ properties: The standard Zicboz extension for cache-block zeroing as ratified in commit 3dd606f ("Create cmobase-v1.0.pdf") of riscv-CMOs.
+ - const: zicfilp + description: | + The standard Zicfilp extension for enforcing forward edge + control-flow integrity as ratified in commit 3f8e450 ("merge + pull request #227 from ved-rivos/0709") of riscv-cfi + github repo. + + - const: zicfiss + description: | + The standard Zicfiss extension for enforcing backward edge + control-flow integrity as ratified in commit 3f8e450 ("merge + pull request #227 from ved-rivos/0709") of riscv-cfi + github repo. + - const: zicntr description: The standard Zicntr extension for base counters and timers, as
This patch adds support for detecting zicfiss and zicfilp. zicfiss and zicfilp stands for unprivleged integer spec extension for shadow stack and branch tracking on indirect branches, respectively.
This patch looks for zicfiss and zicfilp in device tree and accordinlgy lights up bit in cpu feature bitmap. Furthermore this patch adds detection utility functions to return whether shadow stack or landing pads are supported by cpu.
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/cpufeature.h | 13 +++++++++++++ arch/riscv/include/asm/hwcap.h | 2 ++ arch/riscv/include/asm/processor.h | 1 + arch/riscv/kernel/cpufeature.c | 13 +++++++++++++ 4 files changed, 29 insertions(+)
diff --git a/arch/riscv/include/asm/cpufeature.h b/arch/riscv/include/asm/cpufeature.h index 569140d6e639..69007b8100ca 100644 --- a/arch/riscv/include/asm/cpufeature.h +++ b/arch/riscv/include/asm/cpufeature.h @@ -12,6 +12,7 @@ #include <linux/kconfig.h> #include <linux/percpu-defs.h> #include <linux/threads.h> +#include <linux/smp.h> #include <asm/hwcap.h> #include <asm/cpufeature-macros.h>
@@ -137,4 +138,16 @@ static __always_inline bool riscv_cpu_has_extension_unlikely(int cpu, const unsi return __riscv_isa_extension_available(hart_isa[cpu].isa, ext); }
+static inline bool cpu_supports_shadow_stack(void) +{ + return (IS_ENABLED(CONFIG_RISCV_USER_CFI) && + riscv_cpu_has_extension_unlikely(smp_processor_id(), RISCV_ISA_EXT_ZICFISS)); +} + +static inline bool cpu_supports_indirect_br_lp_instr(void) +{ + return (IS_ENABLED(CONFIG_RISCV_USER_CFI) && + riscv_cpu_has_extension_unlikely(smp_processor_id(), RISCV_ISA_EXT_ZICFILP)); +} + #endif diff --git a/arch/riscv/include/asm/hwcap.h b/arch/riscv/include/asm/hwcap.h index 869da082252a..2dc4232bdb3e 100644 --- a/arch/riscv/include/asm/hwcap.h +++ b/arch/riscv/include/asm/hwcap.h @@ -100,6 +100,8 @@ #define RISCV_ISA_EXT_ZICCRSE 91 #define RISCV_ISA_EXT_SVADE 92 #define RISCV_ISA_EXT_SVADU 93 +#define RISCV_ISA_EXT_ZICFILP 94 +#define RISCV_ISA_EXT_ZICFISS 95
#define RISCV_ISA_EXT_XLINUXENVCFG 127
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h index 5f56eb9d114a..e3aba3336e63 100644 --- a/arch/riscv/include/asm/processor.h +++ b/arch/riscv/include/asm/processor.h @@ -13,6 +13,7 @@ #include <vdso/processor.h>
#include <asm/ptrace.h> +#include <asm/hwcap.h>
#define arch_get_mmap_end(addr, len, flags) \ ({ \ diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c index c6ba750536c3..82065cc55822 100644 --- a/arch/riscv/kernel/cpufeature.c +++ b/arch/riscv/kernel/cpufeature.c @@ -150,6 +150,15 @@ static int riscv_ext_svadu_validate(const struct riscv_isa_ext_data *data, return 0; }
+static int riscv_cfi_validate(const struct riscv_isa_ext_data *data, + const unsigned long *isa_bitmap) +{ + if (!IS_ENABLED(CONFIG_RISCV_USER_CFI)) + return -EINVAL; + + return 0; +} + static const unsigned int riscv_zk_bundled_exts[] = { RISCV_ISA_EXT_ZBKB, RISCV_ISA_EXT_ZBKC, @@ -333,6 +342,10 @@ const struct riscv_isa_ext_data riscv_isa_ext[] = { __RISCV_ISA_EXT_SUPERSET_VALIDATE(zicboz, RISCV_ISA_EXT_ZICBOZ, riscv_xlinuxenvcfg_exts, riscv_ext_zicboz_validate), __RISCV_ISA_EXT_DATA(ziccrse, RISCV_ISA_EXT_ZICCRSE), + __RISCV_ISA_EXT_SUPERSET_VALIDATE(zicfilp, RISCV_ISA_EXT_ZICFILP, riscv_xlinuxenvcfg_exts, + riscv_cfi_validate), + __RISCV_ISA_EXT_SUPERSET_VALIDATE(zicfiss, RISCV_ISA_EXT_ZICFISS, riscv_xlinuxenvcfg_exts, + riscv_cfi_validate), __RISCV_ISA_EXT_DATA(zicntr, RISCV_ISA_EXT_ZICNTR), __RISCV_ISA_EXT_DATA(zicond, RISCV_ISA_EXT_ZICOND), __RISCV_ISA_EXT_DATA(zicsr, RISCV_ISA_EXT_ZICSR),
On 14/03/2025 22:39, Deepak Gupta wrote:
I would use riscv_has_extension_unlikely() instead of the cpu specific variant, that would remove the need for #include <linux/smp.h>. Unless you have a good reason to do that?
With the above comment fixed, you can add:
Reviewed-by: Alexandre Ghiti alexghiti@rivosinc.com
Thanks,
Alex
zicfiss and zicfilp extension gets enabled via b3 and b2 in *envcfg CSR. menvcfg controls enabling for S/HS mode. henvcfg control enabling for VS while senvcfg controls enabling for U/VU mode.
zicfilp extension extends *status CSR to hold `expected landing pad` bit. A trap or interrupt can occur between an indirect jmp/call and target instr. `expected landing pad` bit from CPU is recorded into xstatus CSR so that when supervisor performs xret, `expected landing pad` state of CPU can be restored.
zicfiss adds one new CSR - CSR_SSP: CSR_SSP contains current shadow stack pointer.
Signed-off-by: Deepak Gupta debug@rivosinc.com Reviewed-by: Charlie Jenkins charlie@rivosinc.com --- arch/riscv/include/asm/csr.h | 16 ++++++++++++++++ 1 file changed, 16 insertions(+)
diff --git a/arch/riscv/include/asm/csr.h b/arch/riscv/include/asm/csr.h index 6fed42e37705..2f49b9663640 100644 --- a/arch/riscv/include/asm/csr.h +++ b/arch/riscv/include/asm/csr.h @@ -18,6 +18,15 @@ #define SR_MPP _AC(0x00001800, UL) /* Previously Machine */ #define SR_SUM _AC(0x00040000, UL) /* Supervisor User Memory Access */
+/* zicfilp landing pad status bit */ +#define SR_SPELP _AC(0x00800000, UL) +#define SR_MPELP _AC(0x020000000000, UL) +#ifdef CONFIG_RISCV_M_MODE +#define SR_ELP SR_MPELP +#else +#define SR_ELP SR_SPELP +#endif + #define SR_FS _AC(0x00006000, UL) /* Floating-point Status */ #define SR_FS_OFF _AC(0x00000000, UL) #define SR_FS_INITIAL _AC(0x00002000, UL) @@ -212,6 +221,8 @@ #define ENVCFG_PMM_PMLEN_16 (_AC(0x3, ULL) << 32) #define ENVCFG_CBZE (_AC(1, UL) << 7) #define ENVCFG_CBCFE (_AC(1, UL) << 6) +#define ENVCFG_LPE (_AC(1, UL) << 2) +#define ENVCFG_SSE (_AC(1, UL) << 3) #define ENVCFG_CBIE_SHIFT 4 #define ENVCFG_CBIE (_AC(0x3, UL) << ENVCFG_CBIE_SHIFT) #define ENVCFG_CBIE_ILL _AC(0x0, UL) @@ -230,6 +241,11 @@ #define SMSTATEEN0_HSENVCFG (_ULL(1) << SMSTATEEN0_HSENVCFG_SHIFT) #define SMSTATEEN0_SSTATEEN0_SHIFT 63 #define SMSTATEEN0_SSTATEEN0 (_ULL(1) << SMSTATEEN0_SSTATEEN0_SHIFT) +/* + * zicfiss user mode csr + * CSR_SSP holds current shadow stack pointer. + */ +#define CSR_SSP 0x011
/* mseccfg bits */ #define MSECCFG_PMM ENVCFG_PMM
Carves out space in arch specific thread struct for cfi status and shadow stack in usermode on riscv.
This patch does following - defines a new structure cfi_status with status bit for cfi feature - defines shadow stack pointer, base and size in cfi_status structure - defines offsets to new member fields in thread in asm-offsets.c - Saves and restore shadow stack pointer on trap entry (U --> S) and exit (S --> U)
Shadow stack save/restore is gated on feature availiblity and implemented using alternative. CSR can be context switched in `switch_to` as well but soon as kernel shadow stack support gets rolled in, shadow stack pointer will need to be switched at trap entry/exit point (much like `sp`). It can be argued that kernel using shadow stack deployment scenario may not be as prevalant as user mode using this feature. But even if there is some minimal deployment of kernel shadow stack, that means that it needs to be supported. And thus save/restore of shadow stack pointer in entry.S instead of in `switch_to.h`.
Reviewed-by: Charlie Jenkins charlie@rivosinc.com Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/processor.h | 1 + arch/riscv/include/asm/thread_info.h | 3 +++ arch/riscv/include/asm/usercfi.h | 24 ++++++++++++++++++++++++ arch/riscv/kernel/asm-offsets.c | 4 ++++ arch/riscv/kernel/entry.S | 26 ++++++++++++++++++++++++++ 5 files changed, 58 insertions(+)
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h index e3aba3336e63..d851bb5c6da0 100644 --- a/arch/riscv/include/asm/processor.h +++ b/arch/riscv/include/asm/processor.h @@ -14,6 +14,7 @@
#include <asm/ptrace.h> #include <asm/hwcap.h> +#include <asm/usercfi.h>
#define arch_get_mmap_end(addr, len, flags) \ ({ \ diff --git a/arch/riscv/include/asm/thread_info.h b/arch/riscv/include/asm/thread_info.h index f5916a70879a..a0cfe00c2ca6 100644 --- a/arch/riscv/include/asm/thread_info.h +++ b/arch/riscv/include/asm/thread_info.h @@ -62,6 +62,9 @@ struct thread_info { long user_sp; /* User stack pointer */ int cpu; unsigned long syscall_work; /* SYSCALL_WORK_ flags */ +#ifdef CONFIG_RISCV_USER_CFI + struct cfi_status user_cfi_state; +#endif #ifdef CONFIG_SHADOW_CALL_STACK void *scs_base; void *scs_sp; diff --git a/arch/riscv/include/asm/usercfi.h b/arch/riscv/include/asm/usercfi.h new file mode 100644 index 000000000000..5f2027c51917 --- /dev/null +++ b/arch/riscv/include/asm/usercfi.h @@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: GPL-2.0 + * Copyright (C) 2024 Rivos, Inc. + * Deepak Gupta debug@rivosinc.com + */ +#ifndef _ASM_RISCV_USERCFI_H +#define _ASM_RISCV_USERCFI_H + +#ifndef __ASSEMBLY__ +#include <linux/types.h> + +#ifdef CONFIG_RISCV_USER_CFI +struct cfi_status { + unsigned long ubcfi_en : 1; /* Enable for backward cfi. */ + unsigned long rsvd : ((sizeof(unsigned long) * 8) - 1); + unsigned long user_shdw_stk; /* Current user shadow stack pointer */ + unsigned long shdw_stk_base; /* Base address of shadow stack */ + unsigned long shdw_stk_size; /* size of shadow stack */ +}; + +#endif /* CONFIG_RISCV_USER_CFI */ + +#endif /* __ASSEMBLY__ */ + +#endif /* _ASM_RISCV_USERCFI_H */ diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c index e89455a6a0e5..0c188aaf3925 100644 --- a/arch/riscv/kernel/asm-offsets.c +++ b/arch/riscv/kernel/asm-offsets.c @@ -50,6 +50,10 @@ void asm_offsets(void) #endif
OFFSET(TASK_TI_CPU_NUM, task_struct, thread_info.cpu); +#ifdef CONFIG_RISCV_USER_CFI + OFFSET(TASK_TI_CFI_STATUS, task_struct, thread_info.user_cfi_state); + OFFSET(TASK_TI_USER_SSP, task_struct, thread_info.user_cfi_state.user_shdw_stk); +#endif OFFSET(TASK_THREAD_F0, task_struct, thread.fstate.f[0]); OFFSET(TASK_THREAD_F1, task_struct, thread.fstate.f[1]); OFFSET(TASK_THREAD_F2, task_struct, thread.fstate.f[2]); diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S index 33a5a9f2a0d4..68c99124ea55 100644 --- a/arch/riscv/kernel/entry.S +++ b/arch/riscv/kernel/entry.S @@ -147,6 +147,20 @@ SYM_CODE_START(handle_exception)
REG_L s0, TASK_TI_USER_SP(tp) csrrc s1, CSR_STATUS, t0 + /* + * If previous mode was U, capture shadow stack pointer and save it away + * Zero CSR_SSP at the same time for sanitization. + */ + ALTERNATIVE("nop; nop; nop; nop", + __stringify( \ + andi s2, s1, SR_SPP; \ + bnez s2, skip_ssp_save; \ + csrrw s2, CSR_SSP, x0; \ + REG_S s2, TASK_TI_USER_SSP(tp); \ + skip_ssp_save:), + 0, + RISCV_ISA_EXT_ZICFISS, + CONFIG_RISCV_USER_CFI) csrr s2, CSR_EPC csrr s3, CSR_TVAL csrr s4, CSR_CAUSE @@ -236,6 +250,18 @@ SYM_CODE_START_NOALIGN(ret_from_exception) * structures again. */ csrw CSR_SCRATCH, tp + + /* + * Going back to U mode, restore shadow stack pointer + */ + ALTERNATIVE("nop; nop", + __stringify( \ + REG_L s3, TASK_TI_USER_SSP(tp); \ + csrw CSR_SSP, s3), + 0, + RISCV_ISA_EXT_ZICFISS, + CONFIG_RISCV_USER_CFI) + 1: #ifdef CONFIG_RISCV_ISA_V_PREEMPTIVE move a0, sp
On 14/03/2025 22:39, Deepak Gupta wrote:
You could use __nops(4) here instead.
Ditto
Apart from the nits above, you can add:
Reviewed-by: Alexandre Ghiti alexghiti@rivosinc.com
Thanks,
Alex
2025-03-14T14:39:24-07:00, Deepak Gupta debug@rivosinc.com:
I don't think it makes sense to put all the data in thread_info. kernel_ssp and user_ssp is more than enough and the rest can comfortably live elsewhere in task_struct.
thread_info is supposed to be as small as possible -- just spanning multiple cache-lines could be noticeable.
(I'd prefer this closer to the user_sp and kernel_sp swap, it's breaking the flow here. We also already know if we've returned from userspace or not even without SR_SPP, but reusing the information might tangle the logic.)
Are we? I think we can be just as well returning back to kernel-space. Similar to how we can enter the exception handler from kernel-space.
Thanks.
On Thu, Apr 10, 2025 at 01:04:39PM +0200, Radim Krčmář wrote:
I can change it to only include only `user_ssp`, base and size.
But before we go there, see below:
$ pahole -C thread_info kbuild/vmlinux struct thread_info { long unsigned int flags; /* 0 8 */ int preempt_count; /* 8 4 */
/* XXX 4 bytes hole, try to pack */
long int kernel_sp; /* 16 8 */ long int user_sp; /* 24 8 */ int cpu; /* 32 4 */
/* XXX 4 bytes hole, try to pack */
long unsigned int syscall_work; /* 40 8 */ struct cfi_status user_cfi_state; /* 48 32 */ /* --- cacheline 1 boundary (64 bytes) was 16 bytes ago --- */ long unsigned int a0; /* 80 8 */ long unsigned int a1; /* 88 8 */ long unsigned int a2; /* 96 8 */
/* size: 104, cachelines: 2, members: 10 */ /* sum members: 96, holes: 2, sum holes: 8 */ /* last cacheline: 40 bytes */ };
If we were to remove entire `cfi_status`, it would still be 72 bytes (88 bytes if shadow call stack were enabled) and already spans across two cachelines. I did see the comment above that it should fit inside a cacheline. Although I assumed its stale comment given that it already spans across cacheline and I didn't see any special mention in commit messages of changes which grew this structure above one cacheline. So I assumed this was a stale comment.
On the other hand, whenever enable/lock bits are checked, there is a high likelyhood that user_ssp and other fields are going to be accessed and thus it actually might be helpful to have it all in one cacheline during runtime.
So I am not sure if its helpful sticking to the comment which already is stale.
2025-04-23T17:00:29-07:00, Deepak Gupta debug@rivosinc.com:
No need for base and size either -- we don't touch that in the common exception code.
It has only 64 bytes of data without shadow call stack, but it wasted 8 bytes on the holes. a2 is somewhat an outlier that is not used most exception paths and excluding it makes everything fit nicely even now.
Yes, although accessing enable/lock bits will be relatively rare. It seems better to have the overhead during thread setup, rather than on every trap.
So I am not sure if its helpful sticking to the comment which already is stale.
We could fix the holes and also use sp instead of a0 in the new_vmalloc_check, so everything would fit better.
We are really close to fitting into a single cache-line, so I'd prefer if shadow stack only filled thread_info with data that is used very often in the exception handling code.
I think we could do without user_sp in thread_info as well, so there are other packing options.
Btw. could ssp be added to pt_regs?
Thanks.
On Thu, Apr 24, 2025 at 01:52:43PM +0200, Radim Krčmář wrote:
got it.
But we can't exclude shadow call stack. It'll lead to increased size if that config is selected. A solution has to work for all the cases and not half hearted effort.
I don't get what's the big deal if it results in two cachelines. We can (re)organize data structure in a way the most frequently accessed members are together in a single cacheline. We just need to find those members.
In the hot path of exception handling, I see accesses to pt_regs on stack as well. These are definitley different cacheline than thread_info.
I understand the argument of one member field crossing into two cachelines can have undesired perf effects. I do not understand reasoning that thread_info exactly has to fit inside one cacheline.
If this was always supposed to fit in a single cacheline, clearly this invariant isn't/wasn't maintained as changes trickled in. I would like to see what maintainers have to say or someone who did data analysis on this.
I think we could do without user_sp in thread_info as well, so there are other packing options.
Sure, probably somewhere in task_struct. But fact of the matter is that it has to be saved/restore during exception entry/exit. But then load/store to task_struct is essentially a different cachline. Not sure what we will achieve here?
Btw. could ssp be added to pt_regs?
I had that earlier. It breaks user abi. And it was a no go.
Thanks.
2025-04-24T10:56:34-07:00, Deepak Gupta debug@rivosinc.com:
We could drop a0 or user_sp and place the two ints next to each other, saving at least 16 bytes.
(user_sp, a0, a1, and a2 are just temporary storage. I think would be fine with just two temporaries + kernel_sp, to provide three registers for new_vmalloc_check and we never need more.)
Yes, and because this patch is reorganizing the structure, I thought it would be better to do the analysis now, rather than to incur additional debt.
thread_info members are accessed during the first instructions after a trap. We want to maximize the chance that the execution doesn't stall until uarch has time to engage its crystal ball.
In the hot path of exception handling, I see accesses to pt_regs on stack as well. These are definitley different cacheline than thread_info.
Right, and we also access cache-lines for the code.
I don't know how well each uarch keeps the early trap data/code in caches, but it doesn't seem like a bad idea to minimize the amount of cache-lines that are accessed early after trap.
I agree that we could probably lift the constraint for some values -- it's a lot of performance modeling and convincing, though...
In this series, I think it would be good to avoid splitting kernel_sp and a0/a1 into two cache-lines. kernel_sp and a0/a1 are accessed within the first few instructions.
I don't think it is necessary to fix the rest, just not making things worse is already great.
user_sp is only temporarily storage space in thread_info. The sp register is restored from pt_regs, so we could refactor the code to drop user_sp from thread_info.
e.g. use a0, a1, or a2 for the temporary storage: user_sp is not even the userspace sp, it is sp of the previous sp "user", which might have been the kernel.
Btw. could ssp be added to pt_regs?
I had that earlier. It breaks user abi. And it was a no go.
Thanks, I was afraid of that. :)
We might want to eventually push ssp to the stack to follow the same design for trap nesting as sp has, but that can happen when implementing ssp for the kernel. Squeezing into thread_info should work for now.
Sorry forgot to respond to rest of your comments on this thread.
On Thu, Apr 10, 2025 at 01:04:39PM +0200, Radim Krčmář wrote:
<... snipped ...>
If CSR_SCRATCH was 0, then we would be coming from kernel else flow goes to `.Lsave_context`. If we were coming from kernel mode, then eventually flow merges to `.Lsave_context`.
So we will be saving CSR_SSP on all kernel -- > kernel trap handling. That would be unnecessary. IIRC, this was one of the first review comments in early RFC series of these patch series (to not touch CSR_SSP un-necessarily)
We can avoid that by ensuring when we branch by determining if we are coming from user to something like `.Lsave_ssp` which eventually merges into ".Lsave_context". And if we were coming from kernel then we would branch to `.Lsave_context` and thus skipping ssp save logic. But # of branches it introduces in early exception handling is equivalent to what current patches do. So I don't see any value in doing that.
Let me know if I am missing something.
I can remove my comment because it's obvious.
Are we? I think we can be just as well returning back to kernel-space. Similar to how we can enter the exception handler from kernel-space.
Yes we are. See excerpt from `ret_from_exception` in `entry.S`
""" SYM_CODE_START_NOALIGN(ret_from_exception) REG_L s0, PT_STATUS(sp) #ifdef CONFIG_RISCV_M_MODE /* the MPP value is too large to be used as an immediate arg for addi */ li t0, SR_MPP and s0, s0, t0 #else andi s0, s0, SR_SPP #endif bnez s0, 1f
<... snipped ...>
/* * Going back to U mode, restore shadow stack pointer */ ALTERNATIVE("nops(2)", __stringify( \ REG_L s3, TASK_TI_USER_SSP(tp); \ csrw CSR_SSP, s3), 0, RISCV_ISA_EXT_ZICFISS, CONFIG_RISCV_USER_CFI)
1: #ifdef CONFIG_RISCV_ISA_V_PREEMPTIVE move a0, sp call riscv_v_context_nesting_end
<... snipped ...>
"""
2025-04-23T17:23:56-07:00, Deepak Gupta debug@rivosinc.com:
Right, it's hard to avoid the extra branches.
I think we could modify the entry point (STVEC), so we start at different paths based on kernel/userspace trap and only jump once to the common code, like:
SYM_CODE_START(handle_exception_kernel) /* kernel setup magic */ j handle_exception_common SYM_CODE_START(handle_exception_user) /* userspace setup magic */ handle_exception_common:
This is not a suggestion for this series. I would be perfectly happy with just a cleaner code.
Would it be possible to hide the ALTERNATIVE ugliness behind a macro and move it outside the code block that saves pt_regs?
Thanks.
On Thu, Apr 24, 2025 at 02:16:32PM +0200, Radim Krčmář wrote:
Hmm... This can be done. But then it would require to constantly modify `stvec` When you're going back to user mode, you would have to write `stvec` with addr of `handle_exception_user`. But then you can easily get a NMI. It can become ugly. Needs much more thought and on first glance feels error prone.
Only if we have an extension that allows different trap address depending on mode you're coming from (arm does that, right?, I think x86 FRED also does that)
Sure, I'll do something about it.
Thanks.
2025-04-24T11:03:59-07:00, Deepak Gupta debug@rivosinc.com:
We'd just be writing STVEC instead of SSCRATCH, probably at the very same places. It's possible that some micro-architectures will be disturbed more by writing STVEC than SSCRATCH, though, so it's not an easy change to make.
Yeah, the M-mode Linux adds a lot of fun. I don't see support for the Smrnmi extension, so unlucky NMIs should be fatal even now.
`arch_calc_vm_prot_bits` is implemented on risc-v to return VM_READ | VM_WRITE if PROT_WRITE is specified. Similarly `riscv_sys_mmap` is updated to convert all incoming PROT_WRITE to (PROT_WRITE | PROT_READ). This is to make sure that any existing apps using PROT_WRITE still work.
Earlier `protection_map[VM_WRITE]` used to pick read-write PTE encodings. Now `protection_map[VM_WRITE]` will always pick PAGE_SHADOWSTACK PTE encodings for shadow stack. Above changes ensure that existing apps continue to work because underneath kernel will be picking `protection_map[VM_WRITE|VM_READ]` PTE encodings.
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/mman.h | 25 +++++++++++++++++++++++++ arch/riscv/include/asm/pgtable.h | 1 + arch/riscv/kernel/sys_riscv.c | 10 ++++++++++ arch/riscv/mm/init.c | 2 +- 4 files changed, 37 insertions(+), 1 deletion(-)
diff --git a/arch/riscv/include/asm/mman.h b/arch/riscv/include/asm/mman.h new file mode 100644 index 000000000000..392c9c2d2e78 --- /dev/null +++ b/arch/riscv/include/asm/mman.h @@ -0,0 +1,25 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef __ASM_MMAN_H__ +#define __ASM_MMAN_H__ + +#include <linux/compiler.h> +#include <linux/types.h> +#include <uapi/asm/mman.h> + +static inline unsigned long arch_calc_vm_prot_bits(unsigned long prot, + unsigned long pkey __always_unused) +{ + unsigned long ret = 0; + + /* + * If PROT_WRITE was specified, force it to VM_READ | VM_WRITE. + * Only VM_WRITE means shadow stack. + */ + if (prot & PROT_WRITE) + ret = (VM_READ | VM_WRITE); + return ret; +} + +#define arch_calc_vm_prot_bits(prot, pkey) arch_calc_vm_prot_bits(prot, pkey) + +#endif /* ! __ASM_MMAN_H__ */ diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h index 050fdc49b5ad..8c528cd7347a 100644 --- a/arch/riscv/include/asm/pgtable.h +++ b/arch/riscv/include/asm/pgtable.h @@ -178,6 +178,7 @@ extern struct pt_alloc_ops pt_ops __meminitdata; #define PAGE_READ_EXEC __pgprot(_PAGE_BASE | _PAGE_READ | _PAGE_EXEC) #define PAGE_WRITE_EXEC __pgprot(_PAGE_BASE | _PAGE_READ | \ _PAGE_EXEC | _PAGE_WRITE) +#define PAGE_SHADOWSTACK __pgprot(_PAGE_BASE | _PAGE_WRITE)
#define PAGE_COPY PAGE_READ #define PAGE_COPY_EXEC PAGE_READ_EXEC diff --git a/arch/riscv/kernel/sys_riscv.c b/arch/riscv/kernel/sys_riscv.c index d77afe05578f..43a448bf254b 100644 --- a/arch/riscv/kernel/sys_riscv.c +++ b/arch/riscv/kernel/sys_riscv.c @@ -7,6 +7,7 @@
#include <linux/syscalls.h> #include <asm/cacheflush.h> +#include <asm-generic/mman-common.h>
static long riscv_sys_mmap(unsigned long addr, unsigned long len, unsigned long prot, unsigned long flags, @@ -16,6 +17,15 @@ static long riscv_sys_mmap(unsigned long addr, unsigned long len, if (unlikely(offset & (~PAGE_MASK >> page_shift_offset))) return -EINVAL;
+ /* + * If PROT_WRITE is specified then extend that to PROT_READ + * protection_map[VM_WRITE] is now going to select shadow stack encodings. + * So specifying PROT_WRITE actually should select protection_map [VM_WRITE | VM_READ] + * If user wants to create shadow stack then they should use `map_shadow_stack` syscall. + */ + if (unlikely((prot & PROT_WRITE) && !(prot & PROT_READ))) + prot |= PROT_READ; + return ksys_mmap_pgoff(addr, len, prot, flags, fd, offset >> (PAGE_SHIFT - page_shift_offset)); } diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c index 15b2eda4c364..9d6661638d0b 100644 --- a/arch/riscv/mm/init.c +++ b/arch/riscv/mm/init.c @@ -342,7 +342,7 @@ pgd_t early_pg_dir[PTRS_PER_PGD] __initdata __aligned(PAGE_SIZE); static const pgprot_t protection_map[16] = { [VM_NONE] = PAGE_NONE, [VM_READ] = PAGE_READ, - [VM_WRITE] = PAGE_COPY, + [VM_WRITE] = PAGE_SHADOWSTACK, [VM_WRITE | VM_READ] = PAGE_COPY, [VM_EXEC] = PAGE_EXEC, [VM_EXEC | VM_READ] = PAGE_READ_EXEC,
2025-03-14T14:39:25-07:00, Deepak Gupta debug@rivosinc.com:
This function also changes PROT_WX to VM_RWX, which is effectively not changing anything, but I think it deserves an explicit intent. (At least in the commit message.)
Why isn't the previous hunk be enough? (Or why don't we do just this?)
riscv_sys_mmap() eventually calls arch_calc_vm_prot_bits(), so I'd rather fix each code path just once.
Thanks.
On Thu, Apr 10, 2025 at 12:03:44PM +0200, Radim Krčmář wrote:
Will do that.
You're right. Above hunk (arch/riscv/include/asm/mman.h) alone should be enough. I did this change in `sys_riscv.c` out of caution. If it feels like un-necessary, I'll remove it. No hard feelings either way.
Thanks.
2025-04-23T17:45:53-07:00, Deepak Gupta debug@rivosinc.com:
I think it makes the code harder to reason about. Here it is not clear why this caller of ksys_mmap_pgoff() has to do this, while others don't.
On Thu, Apr 24, 2025, at 14:23, Radim Krčmář wrote:
Right, I've been meaning to clean this up for years: there should really not be anything architecture specific in the sys_mmap() or sys_mmap2() syscalls, but we never managed to have a global definition for those and instead leave them up to the architectures and over the years they have diverged.
See below for my earlier prototype of this.
Arnd
From 42efb4468e119f34361f039d0779b8cd00d73de8 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann arnd@arndb.de Date: Wed, 22 Dec 2021 15:06:58 +0100 Subject: [PATCH] treewide: replace mmap_pgoff() with mmap2()
32-bit architectures generally implement sys_mmap2() as their primary mmap() implementation, but common code only provides sys_mmap_pgoff(), which has slightly different behavior when dealing with page sizes other than 4096 bytes.
On m68k, arc and hexagon, the use of mmap_pgoff() appears to be irreversible, as they do support multiple page sizes and user space now relies on passing the argument in actual pages rather than fixed-size units. m68k also has a misleading comment about SUN3 that needs to be rewritten to reflect that this is not broken but required by libc now.
powerpc and riscv are additional exceptions here, as they have checks for the 'prot' argument that need to be kept in place.
Signed-off-by: Arnd Bergmann arnd@arndb.de --- arch/arc/include/asm/syscalls.h | 4 ++++ arch/arm/include/asm/ftrace.h | 4 +--- arch/arm/kernel/entry-common.S | 9 --------- arch/arm/mm/mmap.c | 9 +++++++++ arch/arm64/kernel/sys32.c | 16 ---------------- arch/arm64/tools/syscall_32.tbl | 2 +- arch/csky/kernel/syscall.c | 15 --------------- arch/hexagon/kernel/syscalltab.c | 11 +++++++++-- arch/m68k/include/asm/syscalls.h | 3 +++ arch/m68k/kernel/sys_m68k.c | 11 +++++------ arch/m68k/kernel/syscalltable.S | 2 -- arch/microblaze/kernel/sys_microblaze.c | 11 ----------- arch/mips/kernel/syscall.c | 11 ----------- arch/mips/kernel/syscalls/syscall_o32.tbl | 2 +- arch/nios2/kernel/syscall_table.c | 2 -- arch/parisc/kernel/sys_parisc.c | 10 ---------- arch/powerpc/include/asm/syscalls.h | 7 ++----- arch/powerpc/kernel/syscalls.c | 17 +++++------------ arch/powerpc/kernel/syscalls/syscall.tbl | 4 ++-- arch/riscv/include/asm/syscall.h | 8 ++++++++ arch/riscv/kernel/sys_riscv.c | 5 +++-- arch/riscv/kernel/syscall_table.c | 3 +++ arch/sh/include/asm/syscalls.h | 3 --- arch/sh/kernel/sys_sh.c | 16 ---------------- arch/sparc/kernel/sys_sparc_32.c | 12 ------------ arch/x86/entry/syscalls/syscall_32.tbl | 2 +- arch/xtensa/kernel/syscalls/syscall.tbl | 2 +- include/linux/syscalls.h | 2 +- mm/mmap.c | 11 ++++++++--- mm/nommu.c | 11 ++++++++---
diff --git a/arch/arc/include/asm/syscalls.h b/arch/arc/include/asm/syscalls.h index c3f4714a4f5c..91e2d3fac623 100644 --- a/arch/arc/include/asm/syscalls.h +++ b/arch/arc/include/asm/syscalls.h @@ -17,6 +17,10 @@ int sys_arc_settls(void *); int sys_arc_gettls(void); int sys_arc_usr_cmpxchg(int *, int, int);
+asmlinkage long sys_mmap_pgoff(unsigned long addr, unsigned long len, + unsigned long prot, unsigned long flags, + unsigned long fd, unsigned long pgoff); + #include <asm-generic/syscalls.h>
#endif diff --git a/arch/arm/include/asm/ftrace.h b/arch/arm/include/asm/ftrace.h index 5be3ddc96a50..3b2d84ec11f3 100644 --- a/arch/arm/include/asm/ftrace.h +++ b/arch/arm/include/asm/ftrace.h @@ -62,9 +62,7 @@ static inline void *return_address(unsigned int level) static inline bool arch_syscall_match_sym_name(const char *sym, const char *name) { - if (!strcmp(sym, "sys_mmap2")) - sym = "sys_mmap_pgoff"; - else if (!strcmp(sym, "sys_statfs64_wrapper")) + if (!strcmp(sym, "sys_statfs64_wrapper")) sym = "sys_statfs64"; else if (!strcmp(sym, "sys_fstatfs64_wrapper")) sym = "sys_fstatfs64"; diff --git a/arch/arm/kernel/entry-common.S b/arch/arm/kernel/entry-common.S index f379c852dcb7..d7657d210c61 100644 --- a/arch/arm/kernel/entry-common.S +++ b/arch/arm/kernel/entry-common.S @@ -406,15 +406,6 @@ sys_fstatfs64_wrapper: b sys_fstatfs64 ENDPROC(sys_fstatfs64_wrapper)
-/* - * Note: off_4k (r5) is always units of 4K. If we can't do the requested - * offset, we return EINVAL. - */ -sys_mmap2: - str r5, [sp, #4] - b sys_mmap_pgoff -ENDPROC(sys_mmap2) - #ifdef CONFIG_OABI_COMPAT
/* diff --git a/arch/arm/mm/mmap.c b/arch/arm/mm/mmap.c index 3dbb383c26d5..8d7fb2df1349 100644 --- a/arch/arm/mm/mmap.c +++ b/arch/arm/mm/mmap.c @@ -11,6 +11,7 @@ #include <linux/io.h> #include <linux/personality.h> #include <linux/random.h> +#include <linux/syscalls.h> #include <asm/cachetype.h>
#define COLOUR_ALIGN(addr,pgoff) \ @@ -165,3 +166,11 @@ int valid_mmap_phys_addr_range(unsigned long pfn, size_t size) { return (pfn + (size >> PAGE_SHIFT)) <= (1 + (PHYS_MASK >> PAGE_SHIFT)); } + +/* arc glibc passes PAGE_SIZE units rather than the usual 4K */ +SYSCALL_DEFINE6(mmap_pgoff, unsigned long, addr, unsigned long, len, + long, prot, unsigned long, flags, + long, fd, unsigned long, pgoff) +{ + return ksys_mmap_pgoff(addr, len, prot, flags, fd, pgoff); +} diff --git a/arch/arm64/kernel/sys32.c b/arch/arm64/kernel/sys32.c index 0e3aa4d762a3..8709521ef7bf 100644 --- a/arch/arm64/kernel/sys32.c +++ b/arch/arm64/kernel/sys32.c @@ -43,22 +43,6 @@ COMPAT_SYSCALL_DEFINE3(aarch32_fstatfs64, unsigned int, fd, compat_size_t, sz, return kcompat_sys_fstatfs64(fd, sz, buf); }
-/* - * Note: off_4k is always in units of 4K. If we can't do the - * requested offset because it is not page-aligned, we return -EINVAL. - */ -COMPAT_SYSCALL_DEFINE6(aarch32_mmap2, unsigned long, addr, unsigned long, len, - unsigned long, prot, unsigned long, flags, - unsigned long, fd, unsigned long, off_4k) -{ - if (off_4k & (~PAGE_MASK >> 12)) - return -EINVAL; - - off_4k >>= (PAGE_SHIFT - 12); - - return ksys_mmap_pgoff(addr, len, prot, flags, fd, off_4k); -} - #ifdef CONFIG_CPU_BIG_ENDIAN #define arg_u32p(name) u32, name##_hi, u32, name##_lo #else diff --git a/arch/arm64/tools/syscall_32.tbl b/arch/arm64/tools/syscall_32.tbl index 0765b3a8d6d6..893a33b94c74 100644 --- a/arch/arm64/tools/syscall_32.tbl +++ b/arch/arm64/tools/syscall_32.tbl @@ -204,7 +204,7 @@ 190 common vfork sys_vfork # SuS compliant getrlimit 191 common ugetrlimit sys_getrlimit compat_sys_getrlimit -192 common mmap2 sys_mmap2 compat_sys_aarch32_mmap2 +192 common mmap2 sys_mmap2 193 common truncate64 sys_truncate64 compat_sys_aarch32_truncate64 194 common ftruncate64 sys_ftruncate64 compat_sys_aarch32_ftruncate64 195 common stat64 sys_stat64 diff --git a/arch/csky/kernel/syscall.c b/arch/csky/kernel/syscall.c index 4540a271ee39..ecde99520fd1 100644 --- a/arch/csky/kernel/syscall.c +++ b/arch/csky/kernel/syscall.c @@ -14,21 +14,6 @@ SYSCALL_DEFINE1(set_thread_area, unsigned long, addr) return 0; }
-SYSCALL_DEFINE6(mmap2, - unsigned long, addr, - unsigned long, len, - unsigned long, prot, - unsigned long, flags, - unsigned long, fd, - unsigned long, offset) -{ - if (unlikely(offset & (~PAGE_MASK >> 12))) - return -EINVAL; - - return ksys_mmap_pgoff(addr, len, prot, flags, fd, - offset >> (PAGE_SHIFT - 12)); -} - /* * for abiv1 the 64bits args should be even th, So we need mov the advice * forward. diff --git a/arch/hexagon/kernel/syscalltab.c b/arch/hexagon/kernel/syscalltab.c index b53e2eead4ac..595ab60c9d7b 100644 --- a/arch/hexagon/kernel/syscalltab.c +++ b/arch/hexagon/kernel/syscalltab.c @@ -14,8 +14,6 @@ #define __SYSCALL(nr, call) [nr] = (call), #define __SYSCALL_WITH_COMPAT(nr, native, compat) __SYSCALL(nr, native)
-#define sys_mmap2 sys_mmap_pgoff - SYSCALL_DEFINE6(hexagon_fadvise64_64, int, fd, int, advice, SC_ARG64(offset), SC_ARG64(len)) { @@ -25,6 +23,15 @@ SYSCALL_DEFINE6(hexagon_fadvise64_64, int, fd, int, advice,
#define sys_sync_file_range sys_sync_file_range2
+/* hexagon libc passes PAGE_SIZE units rather than the usual 4K */ +#define sys_mmap2 sys_mmap_pgoff +SYSCALL_DEFINE6(mmap_pgoff, unsigned long, addr, unsigned long, len, + long, prot, unsigned long, flags, + long, fd, unsigned long, pgoff) +{ + return ksys_mmap_pgoff(addr, len, prot, flags, fd, pgoff); +} + void *sys_call_table[__NR_syscalls] = { #include <asm/syscall_table_32.h> }; diff --git a/arch/m68k/include/asm/syscalls.h b/arch/m68k/include/asm/syscalls.h index fb3639acd07b..70c5f01d8036 100644 --- a/arch/m68k/include/asm/syscalls.h +++ b/arch/m68k/include/asm/syscalls.h @@ -5,6 +5,9 @@ #include <linux/compiler_types.h> #include <linux/linkage.h>
+asmlinkage long sys_mmap_pgoff(unsigned long addr, unsigned long len, + unsigned long prot, unsigned long flags, + unsigned long fd, unsigned long pgoff); asmlinkage int sys_cacheflush(unsigned long addr, int scope, int cache, unsigned long len); asmlinkage int sys_atomic_cmpxchg_32(unsigned long newval, int oldval, int d3, diff --git a/arch/m68k/kernel/sys_m68k.c b/arch/m68k/kernel/sys_m68k.c index 1af5e6082467..87a251d05554 100644 --- a/arch/m68k/kernel/sys_m68k.c +++ b/arch/m68k/kernel/sys_m68k.c @@ -37,15 +37,14 @@
#include "../mm/fault.h"
-asmlinkage long sys_mmap2(unsigned long addr, unsigned long len, +/* + * m68k user space expects to pass units of PAGE_SIZE rather than 4KB, + * so provide a custom wrapper around ksys_mmap_pgoff() + */ +asmlinkage long sys_mmap_pgoff(unsigned long addr, unsigned long len, unsigned long prot, unsigned long flags, unsigned long fd, unsigned long pgoff) { - /* - * This is wrong for sun3 - there PAGE_SIZE is 8Kb, - * so we need to shift the argument down by 1; m68k mmap64(3) - * (in libc) expects the last argument of mmap2 in 4Kb units. - */ return ksys_mmap_pgoff(addr, len, prot, flags, fd, pgoff); }
diff --git a/arch/m68k/kernel/syscalltable.S b/arch/m68k/kernel/syscalltable.S index e25ef4a9df30..d7b8f4840eb4 100644 --- a/arch/m68k/kernel/syscalltable.S +++ b/arch/m68k/kernel/syscalltable.S @@ -14,9 +14,7 @@
#include <linux/linkage.h>
-#ifndef CONFIG_MMU #define sys_mmap2 sys_mmap_pgoff -#endif
#define __SYSCALL(nr, entry) .long entry .section .rodata diff --git a/arch/microblaze/kernel/sys_microblaze.c b/arch/microblaze/kernel/sys_microblaze.c index 0850b099f300..7fde269b5ef3 100644 --- a/arch/microblaze/kernel/sys_microblaze.c +++ b/arch/microblaze/kernel/sys_microblaze.c @@ -42,14 +42,3 @@ SYSCALL_DEFINE6(mmap, unsigned long, addr, unsigned long, len,
return ksys_mmap_pgoff(addr, len, prot, flags, fd, pgoff >> PAGE_SHIFT); } - -SYSCALL_DEFINE6(mmap2, unsigned long, addr, unsigned long, len, - unsigned long, prot, unsigned long, flags, unsigned long, fd, - unsigned long, pgoff) -{ - if (pgoff & (~PAGE_MASK >> 12)) - return -EINVAL; - - return ksys_mmap_pgoff(addr, len, prot, flags, fd, - pgoff >> (PAGE_SHIFT - 12)); -} diff --git a/arch/mips/kernel/syscall.c b/arch/mips/kernel/syscall.c index 1bfc34a2e5b3..f050b7cd6dc0 100644 --- a/arch/mips/kernel/syscall.c +++ b/arch/mips/kernel/syscall.c @@ -69,17 +69,6 @@ SYSCALL_DEFINE6(mips_mmap, unsigned long, addr, unsigned long, len, offset >> PAGE_SHIFT); }
-SYSCALL_DEFINE6(mips_mmap2, unsigned long, addr, unsigned long, len, - unsigned long, prot, unsigned long, flags, unsigned long, fd, - unsigned long, pgoff) -{ - if (pgoff & (~PAGE_MASK >> 12)) - return -EINVAL; - - return ksys_mmap_pgoff(addr, len, prot, flags, fd, - pgoff >> (PAGE_SHIFT - 12)); -} - save_static_function(sys_fork); save_static_function(sys_clone); save_static_function(sys_clone3); diff --git a/arch/mips/kernel/syscalls/syscall_o32.tbl b/arch/mips/kernel/syscalls/syscall_o32.tbl index 114a5a1a6230..961379b9ef7b 100644 --- a/arch/mips/kernel/syscalls/syscall_o32.tbl +++ b/arch/mips/kernel/syscalls/syscall_o32.tbl @@ -221,7 +221,7 @@ 207 o32 sendfile sys_sendfile compat_sys_sendfile 208 o32 getpmsg sys_ni_syscall 209 o32 putpmsg sys_ni_syscall -210 o32 mmap2 sys_mips_mmap2 +210 o32 mmap2 sys_mmap2 211 o32 truncate64 sys_truncate64 sys_32_truncate64 212 o32 ftruncate64 sys_ftruncate64 sys_32_ftruncate64 213 o32 stat64 sys_stat64 sys_newstat diff --git a/arch/nios2/kernel/syscall_table.c b/arch/nios2/kernel/syscall_table.c index 434694067d8f..88ecde1b96f3 100644 --- a/arch/nios2/kernel/syscall_table.c +++ b/arch/nios2/kernel/syscall_table.c @@ -12,8 +12,6 @@ #define __SYSCALL(nr, call) [nr] = (call), #define __SYSCALL_WITH_COMPAT(nr, native, compat) __SYSCALL(nr, native)
-#define sys_mmap2 sys_mmap_pgoff - void *sys_call_table[__NR_syscalls] = { [0 ... __NR_syscalls-1] = sys_ni_syscall, #include <asm/syscall_table_32.h> diff --git a/arch/parisc/kernel/sys_parisc.c b/arch/parisc/kernel/sys_parisc.c index f852fe274abe..c56d4d3dff32 100644 --- a/arch/parisc/kernel/sys_parisc.c +++ b/arch/parisc/kernel/sys_parisc.c @@ -182,16 +182,6 @@ unsigned long arch_get_unmapped_area_topdown(struct file *filp, addr, len, pgoff, flags, DOWN); }
-asmlinkage unsigned long sys_mmap2(unsigned long addr, unsigned long len, - unsigned long prot, unsigned long flags, unsigned long fd, - unsigned long pgoff) -{ - /* Make sure the shift for mmap2 is constant (12), no matter what PAGE_SIZE - we have. */ - return ksys_mmap_pgoff(addr, len, prot, flags, fd, - pgoff >> (PAGE_SHIFT - 12)); -} - asmlinkage unsigned long sys_mmap(unsigned long addr, unsigned long len, unsigned long prot, unsigned long flags, unsigned long fd, unsigned long offset) diff --git a/arch/powerpc/include/asm/syscalls.h b/arch/powerpc/include/asm/syscalls.h index 6d51b007b59e..711785d630c2 100644 --- a/arch/powerpc/include/asm/syscalls.h +++ b/arch/powerpc/include/asm/syscalls.h @@ -52,10 +52,10 @@ long compat_sys_ppc64_personality(unsigned long personality);
long sys_swapcontext(struct ucontext __user *old_ctx, struct ucontext __user *new_ctx, long ctx_size); -long sys_mmap(unsigned long addr, size_t len, +long sys_ppc_mmap(unsigned long addr, size_t len, unsigned long prot, unsigned long flags, unsigned long fd, off_t offset); -long sys_mmap2(unsigned long addr, size_t len, +long sys_ppc32_mmap2(unsigned long addr, size_t len, unsigned long prot, unsigned long flags, unsigned long fd, unsigned long pgoff); long sys_switch_endian(void); @@ -113,9 +113,6 @@ long sys_ppc_fallocate(int fd, int mode, u32 offset1, u32 offset2, u32 len1, u32 len2); #endif #ifdef CONFIG_COMPAT -long compat_sys_mmap2(unsigned long addr, size_t len, - unsigned long prot, unsigned long flags, - unsigned long fd, unsigned long pgoff); long compat_sys_ppc_pread64(unsigned int fd, char __user *ubuf, compat_size_t count, u32 reg6, u32 pos1, u32 pos2); diff --git a/arch/powerpc/kernel/syscalls.c b/arch/powerpc/kernel/syscalls.c index 68ebb23a5af4..003ad44eaa5f 100644 --- a/arch/powerpc/kernel/syscalls.c +++ b/arch/powerpc/kernel/syscalls.c @@ -49,24 +49,17 @@ static long do_mmap2(unsigned long addr, size_t len, return ksys_mmap_pgoff(addr, len, prot, flags, fd, off >> shift); }
-SYSCALL_DEFINE6(mmap2, unsigned long, addr, size_t, len, +/* + * like the normal mmap2(), but checks the prot argument + */ +SYSCALL_DEFINE6(ppc32_mmap2, unsigned long, addr, size_t, len, unsigned long, prot, unsigned long, flags, unsigned long, fd, unsigned long, pgoff) { return do_mmap2(addr, len, prot, flags, fd, pgoff, PAGE_SHIFT-12); }
-#ifdef CONFIG_COMPAT -COMPAT_SYSCALL_DEFINE6(mmap2, - unsigned long, addr, size_t, len, - unsigned long, prot, unsigned long, flags, - unsigned long, fd, unsigned long, off_4k) -{ - return do_mmap2(addr, len, prot, flags, fd, off_4k, PAGE_SHIFT-12); -} -#endif - -SYSCALL_DEFINE6(mmap, unsigned long, addr, size_t, len, +SYSCALL_DEFINE6(ppc_mmap, unsigned long, addr, size_t, len, unsigned long, prot, unsigned long, flags, unsigned long, fd, off_t, offset) { diff --git a/arch/powerpc/kernel/syscalls/syscall.tbl b/arch/powerpc/kernel/syscalls/syscall.tbl index 9a084bdb8926..b7d83806d48f 100644 --- a/arch/powerpc/kernel/syscalls/syscall.tbl +++ b/arch/powerpc/kernel/syscalls/syscall.tbl @@ -124,7 +124,7 @@ 89 32 readdir sys_old_readdir compat_sys_old_readdir 89 64 readdir sys_ni_syscall 89 spu readdir sys_ni_syscall -90 common mmap sys_mmap +90 common mmap sys_ppc_mmap 91 common munmap sys_munmap 92 common truncate sys_truncate compat_sys_truncate 93 common ftruncate sys_ftruncate compat_sys_ftruncate @@ -249,7 +249,7 @@ 191 32 readahead sys_ppc_readahead compat_sys_ppc_readahead 191 64 readahead sys_readahead 191 spu readahead sys_readahead -192 32 mmap2 sys_mmap2 compat_sys_mmap2 +192 32 mmap2 sys_mmap2 193 32 truncate64 sys_ppc_truncate64 compat_sys_ppc_truncate64 194 32 ftruncate64 sys_ppc_ftruncate64 compat_sys_ppc_ftruncate64 195 32 stat64 sys_stat64 diff --git a/arch/riscv/include/asm/syscall.h b/arch/riscv/include/asm/syscall.h index 34313387f977..20cbf0c63e20 100644 --- a/arch/riscv/include/asm/syscall.h +++ b/arch/riscv/include/asm/syscall.h @@ -121,4 +121,12 @@ asmlinkage long sys_riscv_flush_icache(uintptr_t, uintptr_t, uintptr_t);
asmlinkage long sys_riscv_hwprobe(struct riscv_hwprobe *, size_t, size_t, unsigned long *, unsigned int); + +asmlinkage long sys_riscv_mmap2(unsigned long addr, unsigned long len, + unsigned long prot, unsigned long flags, + unsigned long fd, unsigned long pgoff); +asmlinkage long sys_riscv_mmap(unsigned long addr, unsigned long len, + unsigned long prot, unsigned long flags, + unsigned long fd, off_t pgoff); + #endif /* _ASM_RISCV_SYSCALL_H */ diff --git a/arch/riscv/kernel/sys_riscv.c b/arch/riscv/kernel/sys_riscv.c index d77afe05578f..8e8c32241db6 100644 --- a/arch/riscv/kernel/sys_riscv.c +++ b/arch/riscv/kernel/sys_riscv.c @@ -20,8 +20,9 @@ static long riscv_sys_mmap(unsigned long addr, unsigned long len, offset >> (PAGE_SHIFT - page_shift_offset)); }
+/* same as the generic version, but checks the prot argument */ #ifdef CONFIG_64BIT -SYSCALL_DEFINE6(mmap, unsigned long, addr, unsigned long, len, +SYSCALL_DEFINE6(riscv_mmap, unsigned long, addr, unsigned long, len, unsigned long, prot, unsigned long, flags, unsigned long, fd, unsigned long, offset) { @@ -30,7 +31,7 @@ SYSCALL_DEFINE6(mmap, unsigned long, addr, unsigned long, len, #endif
#if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT) -SYSCALL_DEFINE6(mmap2, unsigned long, addr, unsigned long, len, +SYSCALL_DEFINE6(riscv_mmap2, unsigned long, addr, unsigned long, len, unsigned long, prot, unsigned long, flags, unsigned long, fd, unsigned long, offset) { diff --git a/arch/riscv/kernel/syscall_table.c b/arch/riscv/kernel/syscall_table.c index 43363a80d722..def8e888d262 100644 --- a/arch/riscv/kernel/syscall_table.c +++ b/arch/riscv/kernel/syscall_table.c @@ -15,6 +15,9 @@ #define __SYSCALL(nr, call) asmlinkage long __riscv_##call(const struct pt_regs *); #include <asm/syscall_table.h>
+#define __riscv_sys_mmap __riscv_sys_riscv_mmap +#define __riscv_sys_mmap2 __riscv_sys_riscv_mmap2 + #undef __SYSCALL #define __SYSCALL(nr, call) [nr] = __riscv_##call,
diff --git a/arch/sh/include/asm/syscalls.h b/arch/sh/include/asm/syscalls.h index 39240e06e8aa..d54c69633acc 100644 --- a/arch/sh/include/asm/syscalls.h +++ b/arch/sh/include/asm/syscalls.h @@ -5,9 +5,6 @@ asmlinkage int old_mmap(unsigned long addr, unsigned long len, unsigned long prot, unsigned long flags, int fd, unsigned long off); -asmlinkage long sys_mmap2(unsigned long addr, unsigned long len, - unsigned long prot, unsigned long flags, - unsigned long fd, unsigned long pgoff); asmlinkage int sys_cacheflush(unsigned long addr, unsigned long len, int op);
#include <asm/syscalls_32.h> diff --git a/arch/sh/kernel/sys_sh.c b/arch/sh/kernel/sys_sh.c index a5a7b33ed81a..45abd5e1246e 100644 --- a/arch/sh/kernel/sys_sh.c +++ b/arch/sh/kernel/sys_sh.c @@ -38,22 +38,6 @@ asmlinkage int old_mmap(unsigned long addr, unsigned long len, return ksys_mmap_pgoff(addr, len, prot, flags, fd, off>>PAGE_SHIFT); }
-asmlinkage long sys_mmap2(unsigned long addr, unsigned long len, - unsigned long prot, unsigned long flags, - unsigned long fd, unsigned long pgoff) -{ - /* - * The shift for mmap2 is constant, regardless of PAGE_SIZE - * setting. - */ - if (pgoff & ((1 << (PAGE_SHIFT - 12)) - 1)) - return -EINVAL; - - pgoff >>= PAGE_SHIFT - 12; - - return ksys_mmap_pgoff(addr, len, prot, flags, fd, pgoff); -} - /* sys_cacheflush -- flush (part of) the processor cache. */ asmlinkage int sys_cacheflush(unsigned long addr, unsigned long len, int op) { diff --git a/arch/sparc/kernel/sys_sparc_32.c b/arch/sparc/kernel/sys_sparc_32.c index fb31bc0c5b48..2331ba5bad8b 100644 --- a/arch/sparc/kernel/sys_sparc_32.c +++ b/arch/sparc/kernel/sys_sparc_32.c @@ -104,18 +104,6 @@ int sparc_mmap_check(unsigned long addr, unsigned long len) return 0; }
-/* Linux version of mmap */ - -SYSCALL_DEFINE6(mmap2, unsigned long, addr, unsigned long, len, - unsigned long, prot, unsigned long, flags, unsigned long, fd, - unsigned long, pgoff) -{ - /* Make sure the shift for mmap2 is constant (12), no matter what PAGE_SIZE - we have. */ - return ksys_mmap_pgoff(addr, len, prot, flags, fd, - pgoff >> (PAGE_SHIFT - 12)); -} - SYSCALL_DEFINE6(mmap, unsigned long, addr, unsigned long, len, unsigned long, prot, unsigned long, flags, unsigned long, fd, unsigned long, off) diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl index ac007ea00979..e11dba577071 100644 --- a/arch/x86/entry/syscalls/syscall_32.tbl +++ b/arch/x86/entry/syscalls/syscall_32.tbl @@ -204,7 +204,7 @@ 189 i386 putpmsg 190 i386 vfork sys_vfork 191 i386 ugetrlimit sys_getrlimit compat_sys_getrlimit -192 i386 mmap2 sys_mmap_pgoff +192 i386 mmap2 sys_mmap2 193 i386 truncate64 sys_ia32_truncate64 194 i386 ftruncate64 sys_ia32_ftruncate64 195 i386 stat64 sys_stat64 compat_sys_ia32_stat64 diff --git a/arch/xtensa/kernel/syscalls/syscall.tbl b/arch/xtensa/kernel/syscalls/syscall.tbl index f657a77314f8..e13e73fab37c 100644 --- a/arch/xtensa/kernel/syscalls/syscall.tbl +++ b/arch/xtensa/kernel/syscalls/syscall.tbl @@ -89,7 +89,7 @@ 78 common flistxattr sys_flistxattr 79 common fremovexattr sys_fremovexattr # File Map / Shared Memory Operations -80 common mmap2 sys_mmap_pgoff +80 common mmap2 sys_mmap2 81 common munmap sys_munmap 82 common mprotect sys_mprotect 83 common brk sys_brk diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h index e5603cc91963..0df21f8f6e24 100644 --- a/include/linux/syscalls.h +++ b/include/linux/syscalls.h @@ -1198,7 +1198,7 @@ asmlinkage long sys_ipc(unsigned int call, int first, unsigned long second, unsigned long third, void __user *ptr, long fifth);
/* obsolete */ -asmlinkage long sys_mmap_pgoff(unsigned long addr, unsigned long len, +asmlinkage long sys_mmap2(unsigned long addr, unsigned long len, unsigned long prot, unsigned long flags, unsigned long fd, unsigned long pgoff); asmlinkage long sys_old_mmap(struct mmap_arg_struct __user *arg); diff --git a/mm/mmap.c b/mm/mmap.c index bd210aaf7ebd..73ba0e5cae07 100644 --- a/mm/mmap.c +++ b/mm/mmap.c @@ -611,11 +611,16 @@ unsigned long ksys_mmap_pgoff(unsigned long addr, unsigned long len, return retval; }
-SYSCALL_DEFINE6(mmap_pgoff, unsigned long, addr, unsigned long, len, +SYSCALL_DEFINE6(mmap2, unsigned long, addr, unsigned long, len, unsigned long, prot, unsigned long, flags, - unsigned long, fd, unsigned long, pgoff) + unsigned long, fd, unsigned long, off_4k) { - return ksys_mmap_pgoff(addr, len, prot, flags, fd, pgoff); + if (off_4k & (~PAGE_MASK >> 12)) + return -EINVAL; + + off_4k >>= (PAGE_SHIFT - 12); + + return ksys_mmap_pgoff(addr, len, prot, flags, fd, off_4k); }
#ifdef __ARCH_WANT_SYS_OLD_MMAP diff --git a/mm/nommu.c b/mm/nommu.c index 76c3096206fa..86cca5b21e87 100644 --- a/mm/nommu.c +++ b/mm/nommu.c @@ -1247,11 +1247,16 @@ unsigned long ksys_mmap_pgoff(unsigned long addr, unsigned long len, return retval; }
-SYSCALL_DEFINE6(mmap_pgoff, unsigned long, addr, unsigned long, len, +SYSCALL_DEFINE6(mmap2, unsigned long, addr, unsigned long, len, unsigned long, prot, unsigned long, flags, - unsigned long, fd, unsigned long, pgoff) + unsigned long, fd, unsigned long, off_4k) { - return ksys_mmap_pgoff(addr, len, prot, flags, fd, pgoff); + if (off_4k & (~PAGE_MASK >> 12)) + return -EINVAL; + + off_4k >>= (PAGE_SHIFT - 12); + + return ksys_mmap_pgoff(addr, len, prot, flags, fd, off_4k); }
#ifdef __ARCH_WANT_SYS_OLD_MMAP diff --git a/tools/perf/arch/powerpc/entry/syscalls/syscall.tbl b/tools/perf/arch/powerpc/entry/syscalls/syscall.tbl index 9a084bdb8926..ddfeff063a6d 100644 --- a/tools/perf/arch/powerpc/entry/syscalls/syscall.tbl +++ b/tools/perf/arch/powerpc/entry/syscalls/syscall.tbl @@ -124,7 +124,7 @@ 89 32 readdir sys_old_readdir compat_sys_old_readdir 89 64 readdir sys_ni_syscall 89 spu readdir sys_ni_syscall -90 common mmap sys_mmap +90 common mmap sys_ppc_mmap 91 common munmap sys_munmap 92 common truncate sys_truncate compat_sys_truncate 93 common ftruncate sys_ftruncate compat_sys_ftruncate @@ -249,7 +249,7 @@ 191 32 readahead sys_ppc_readahead compat_sys_ppc_readahead 191 64 readahead sys_readahead 191 spu readahead sys_readahead -192 32 mmap2 sys_mmap2 compat_sys_mmap2 +192 32 mmap2 sys_ppc32_mmap2 193 32 truncate64 sys_ppc_truncate64 compat_sys_ppc_truncate64 194 32 ftruncate64 sys_ppc_ftruncate64 compat_sys_ppc_ftruncate64 195 32 stat64 sys_stat64
This patch implements creating shadow stack pte (on riscv). Creating shadow stack PTE on riscv means that clearing RWX and then setting W=1.
Reviewed-by: Alexandre Ghiti alexghiti@rivosinc.com Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/pgtable.h | 10 ++++++++++ 1 file changed, 10 insertions(+)
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h index 8c528cd7347a..ede43185ffdf 100644 --- a/arch/riscv/include/asm/pgtable.h +++ b/arch/riscv/include/asm/pgtable.h @@ -421,6 +421,11 @@ static inline pte_t pte_mkwrite_novma(pte_t pte) return __pte(pte_val(pte) | _PAGE_WRITE); }
+static inline pte_t pte_mkwrite_shstk(pte_t pte) +{ + return __pte((pte_val(pte) & ~(_PAGE_LEAF)) | _PAGE_WRITE); +} + /* static inline pte_t pte_mkexec(pte_t pte) */
static inline pte_t pte_mkdirty(pte_t pte) @@ -749,6 +754,11 @@ static inline pmd_t pmd_mkwrite_novma(pmd_t pmd) return pte_pmd(pte_mkwrite_novma(pmd_pte(pmd))); }
+static inline pmd_t pmd_mkwrite_shstk(pmd_t pte) +{ + return __pmd((pmd_val(pte) & ~(_PAGE_LEAF)) | _PAGE_WRITE); +} + static inline pmd_t pmd_wrprotect(pmd_t pmd) { return pte_pmd(pte_wrprotect(pmd_pte(pmd)));
pte_mkwrite creates PTEs with WRITE encodings for underlying arch. Underlying arch can have two types of writeable mappings. One that can be written using regular store instructions. Another one that can only be written using specialized store instructions (like shadow stack stores). pte_mkwrite can select write PTE encoding based on VMA range (i.e. VM_SHADOW_STACK)
Reviewed-by: Alexandre Ghiti alexghiti@rivosinc.com Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/pgtable.h | 7 +++++++ arch/riscv/mm/pgtable.c | 17 +++++++++++++++++ 2 files changed, 24 insertions(+)
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h index ede43185ffdf..ccd2fa34afb8 100644 --- a/arch/riscv/include/asm/pgtable.h +++ b/arch/riscv/include/asm/pgtable.h @@ -416,6 +416,10 @@ static inline pte_t pte_wrprotect(pte_t pte)
/* static inline pte_t pte_mkread(pte_t pte) */
+struct vm_area_struct; +pte_t pte_mkwrite(pte_t pte, struct vm_area_struct *vma); +#define pte_mkwrite pte_mkwrite + static inline pte_t pte_mkwrite_novma(pte_t pte) { return __pte(pte_val(pte) | _PAGE_WRITE); @@ -749,6 +753,9 @@ static inline pmd_t pmd_mkyoung(pmd_t pmd) return pte_pmd(pte_mkyoung(pmd_pte(pmd))); }
+pmd_t pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); +#define pmd_mkwrite pmd_mkwrite + static inline pmd_t pmd_mkwrite_novma(pmd_t pmd) { return pte_pmd(pte_mkwrite_novma(pmd_pte(pmd))); diff --git a/arch/riscv/mm/pgtable.c b/arch/riscv/mm/pgtable.c index 4ae67324f992..be5d38546bb3 100644 --- a/arch/riscv/mm/pgtable.c +++ b/arch/riscv/mm/pgtable.c @@ -155,3 +155,20 @@ pmd_t pmdp_collapse_flush(struct vm_area_struct *vma, return pmd; } #endif /* CONFIG_TRANSPARENT_HUGEPAGE */ + +pte_t pte_mkwrite(pte_t pte, struct vm_area_struct *vma) +{ + if (vma->vm_flags & VM_SHADOW_STACK) + return pte_mkwrite_shstk(pte); + + return pte_mkwrite_novma(pte); +} + +pmd_t pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma) +{ + if (vma->vm_flags & VM_SHADOW_STACK) + return pmd_mkwrite_shstk(pmd); + + return pmd_mkwrite_novma(pmd); +} +
`fork` implements copy on write (COW) by making pages readonly in child and parent both.
ptep_set_wrprotect and pte_wrprotect clears _PAGE_WRITE in PTE. Assumption is that page is readable and on fault copy on write happens.
To implement COW on shadow stack pages, clearing up W bit makes them XWR = 000. This will result in wrong PTE setting which says no perms but V=1 and PFN field pointing to final page. Instead desired behavior is to turn it into a readable page, take an access (load/store) fault on sspush/sspop (shadow stack) and then perform COW on such pages. This way regular reads would still be allowed and not lead to COW maintaining current behavior of COW on non-shadow stack but writeable memory.
On the other hand it doesn't interfere with existing COW for read-write memory. Assumption is always that _PAGE_READ must have been set and thus setting _PAGE_READ is harmless.
Reviewed-by: Alexandre Ghiti alexghiti@rivosinc.com Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/pgtable.h | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h index ccd2fa34afb8..54707686f042 100644 --- a/arch/riscv/include/asm/pgtable.h +++ b/arch/riscv/include/asm/pgtable.h @@ -411,7 +411,7 @@ static inline int pte_devmap(pte_t pte)
static inline pte_t pte_wrprotect(pte_t pte) { - return __pte(pte_val(pte) & ~(_PAGE_WRITE)); + return __pte((pte_val(pte) & ~(_PAGE_WRITE)) | (_PAGE_READ)); }
/* static inline pte_t pte_mkread(pte_t pte) */ @@ -612,7 +612,15 @@ static inline pte_t ptep_get_and_clear(struct mm_struct *mm, static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long address, pte_t *ptep) { - atomic_long_and(~(unsigned long)_PAGE_WRITE, (atomic_long_t *)ptep); + pte_t read_pte = READ_ONCE(*ptep); + /* + * ptep_set_wrprotect can be called for shadow stack ranges too. + * shadow stack memory is XWR = 010 and thus clearing _PAGE_WRITE will lead to + * encoding 000b which is wrong encoding with V = 1. This should lead to page fault + * but we dont want this wrong configuration to be set in page tables. + */ + atomic_long_set((atomic_long_t *)ptep, + ((pte_val(read_pte) & ~(unsigned long)_PAGE_WRITE) | _PAGE_READ)); }
#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
As discussed extensively in the changelog for the addition of this syscall on x86 ("x86/shstk: Introduce map_shadow_stack syscall") the existing mmap() and madvise() syscalls do not map entirely well onto the security requirements for shadow stack memory since they lead to windows where memory is allocated but not yet protected or stacks which are not properly and safely initialised. Instead a new syscall map_shadow_stack() has been defined which allocates and initialises a shadow stack page.
This patch implements this syscall for riscv. riscv doesn't require token to be setup by kernel because user mode can do that by itself. However to provide compatibility and portability with other architectues, user mode can specify token set flag.
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/kernel/Makefile | 1 + arch/riscv/kernel/usercfi.c | 144 ++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 145 insertions(+)
diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile index 8d186bfced45..3a861d320654 100644 --- a/arch/riscv/kernel/Makefile +++ b/arch/riscv/kernel/Makefile @@ -125,3 +125,4 @@ obj-$(CONFIG_ACPI) += acpi.o obj-$(CONFIG_ACPI_NUMA) += acpi_numa.o
obj-$(CONFIG_GENERIC_CPU_VULNERABILITIES) += bugs.o +obj-$(CONFIG_RISCV_USER_CFI) += usercfi.o diff --git a/arch/riscv/kernel/usercfi.c b/arch/riscv/kernel/usercfi.c new file mode 100644 index 000000000000..24022809a7b5 --- /dev/null +++ b/arch/riscv/kernel/usercfi.c @@ -0,0 +1,144 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2024 Rivos, Inc. + * Deepak Gupta debug@rivosinc.com + */ + +#include <linux/sched.h> +#include <linux/bitops.h> +#include <linux/types.h> +#include <linux/mm.h> +#include <linux/mman.h> +#include <linux/uaccess.h> +#include <linux/sizes.h> +#include <linux/user.h> +#include <linux/syscalls.h> +#include <linux/prctl.h> +#include <asm/csr.h> +#include <asm/usercfi.h> + +#define SHSTK_ENTRY_SIZE sizeof(void *) + +/* + * Writes on shadow stack can either be `sspush` or `ssamoswap`. `sspush` can happen + * implicitly on current shadow stack pointed to by CSR_SSP. `ssamoswap` takes pointer to + * shadow stack. To keep it simple, we plan to use `ssamoswap` to perform writes on shadow + * stack. + */ +static noinline unsigned long amo_user_shstk(unsigned long *addr, unsigned long val) +{ + /* + * Never expect -1 on shadow stack. Expect return addresses and zero + */ + unsigned long swap = -1; + + __enable_user_access(); + asm goto( + ".option push\n" + ".option arch, +zicfiss\n" + "1: ssamoswap.d %[swap], %[val], %[addr]\n" + _ASM_EXTABLE(1b, %l[fault]) + RISCV_ACQUIRE_BARRIER + ".option pop\n" + : [swap] "=r" (swap), [addr] "+A" (*addr) + : [val] "r" (val) + : "memory" + : fault + ); + __disable_user_access(); + return swap; +fault: + __disable_user_access(); + return -1; +} + +/* + * Create a restore token on the shadow stack. A token is always XLEN wide + * and aligned to XLEN. + */ +static int create_rstor_token(unsigned long ssp, unsigned long *token_addr) +{ + unsigned long addr; + + /* Token must be aligned */ + if (!IS_ALIGNED(ssp, SHSTK_ENTRY_SIZE)) + return -EINVAL; + + /* On RISC-V we're constructing token to be function of address itself */ + addr = ssp - SHSTK_ENTRY_SIZE; + + if (amo_user_shstk((unsigned long __user *)addr, (unsigned long)ssp) == -1) + return -EFAULT; + + if (token_addr) + *token_addr = addr; + + return 0; +} + +static unsigned long allocate_shadow_stack(unsigned long addr, unsigned long size, + unsigned long token_offset, bool set_tok) +{ + int flags = MAP_ANONYMOUS | MAP_PRIVATE; + struct mm_struct *mm = current->mm; + unsigned long populate, tok_loc = 0; + + if (addr) + flags |= MAP_FIXED_NOREPLACE; + + mmap_write_lock(mm); + addr = do_mmap(NULL, addr, size, PROT_READ, flags, + VM_SHADOW_STACK | VM_WRITE, 0, &populate, NULL); + mmap_write_unlock(mm); + + if (!set_tok || IS_ERR_VALUE(addr)) + goto out; + + if (create_rstor_token(addr + token_offset, &tok_loc)) { + vm_munmap(addr, size); + return -EINVAL; + } + + addr = tok_loc; + +out: + return addr; +} + +SYSCALL_DEFINE3(map_shadow_stack, unsigned long, addr, unsigned long, size, unsigned int, flags) +{ + bool set_tok = flags & SHADOW_STACK_SET_TOKEN; + unsigned long aligned_size = 0; + + if (!cpu_supports_shadow_stack()) + return -EOPNOTSUPP; + + /* Anything other than set token should result in invalid param */ + if (flags & ~SHADOW_STACK_SET_TOKEN) + return -EINVAL; + + /* + * Unlike other architectures, on RISC-V, SSP pointer is held in CSR_SSP and is available + * CSR in all modes. CSR accesses are performed using 12bit index programmed in instruction + * itself. This provides static property on register programming and writes to CSR can't + * be unintentional from programmer's perspective. As long as programmer has guarded areas + * which perform writes to CSR_SSP properly, shadow stack pivoting is not possible. Since + * CSR_SSP is writeable by user mode, it itself can setup a shadow stack token subsequent + * to allocation. Although in order to provide portablity with other architecture (because + * `map_shadow_stack` is arch agnostic syscall), RISC-V will follow expectation of a token + * flag in flags and if provided in flags, setup a token at the base. + */ + + /* If there isn't space for a token */ + if (set_tok && size < SHSTK_ENTRY_SIZE) + return -ENOSPC; + + if (addr && (addr & (PAGE_SIZE - 1))) + return -EINVAL; + + aligned_size = PAGE_ALIGN(size); + if (aligned_size < size) + return -EOVERFLOW; + + return allocate_shadow_stack(addr, aligned_size, size, set_tok); +}
On Sat, Mar 15, 2025 at 5:39 AM Deepak Gupta debug@rivosinc.com wrote:
Hi Deepak, It just came to my mind, do we need to ensure that menvcfg.SSE is not zero before executing the ssamoswap instruction? Since ssamoswap is not encoded using MOP, I’m wondering if we should make sure that executing ssamoswap won’t accidentally trigger an illegal instruction exception. Thanks.
On Mon, Apr 07, 2025 at 12:50:35PM +0800, Zong Li wrote:
FWFT patches turn SSE during early boot. There is a bug there though, I need to check if those FWFT SBI call succeeded or not. If it failed then itshould set a global variable indicating shadow stack can't be turned on. And in that case this flow wouldn't be reachable. Soon I will post v13 with these changes.
Thanks for noticing.
2025-03-14T14:39:29-07:00, Deepak Gupta debug@rivosinc.com:
RISC-V shadow stack could use mmap() and madvise() perfectly well. Userspace can always initialize the shadow stack properly and the shadow stack memory is never protected from other malicious threads.
I think that the compatibility argument is reasonable. We'd need to modify the other syscalls to allow a write-only mapping anyway.
Shouldn't compiler accept ssamoswap.d opcode even without zicfiss arch?
Why is the barrier here?
I think we should return 0 and -EFAULT. We can ignore the swapped value, or return it through a pointer.
Is MAP_GROWSDOWN pointless?
PROT_READ implies VM_READ, so won't this select PAGE_COPY in the protection_map instead of PAGE_SHADOWSTACK?
Wouldn't avoiding VM_READ also allow us to get rid of the ugly hack in pte_mkwrite? (VM_WRITE would naturally select the right XWR flags.)
if (!PAGE_ALIGNED(addr))
On Thu, Apr 10, 2025 at 11:56:44AM +0200, Radim Krčmář wrote:
Deviating from what other arches are doing will create more thrash. I expect there will be merging of common logic between x86, arm64 and riscv. Infact I did post one such RFC patch set last year (didn't follow up on it). Using `mmap/madvise` defeats that purpose of creating common logic between arches.
There are pitfalls as mentioned with respect to mmap/madivse because of unique nature of shadow stack. And thus it was accepted to create a new syscall to create such mappings. RISC-V will stick to that.
Userspace can always initialize the shadow stack properly and the shadow stack memory is never protected from other malicious threads.
Shadow stack memory is protected from inadvertent stores (be it same thread or a different thread in same address space). Malicious code which can do `sspush/ssamoswap` already assumes that code integrity policies are broken.
I think that the compatibility argument is reasonable. We'd need to modify the other syscalls to allow a write-only mapping anyway.
Its illegal instruction if shadow stack aren't available. Current toolchain emits it only if zicfiss is specified in march.
IIRC, I was following `arch_cmpxchg_acquire`. But I think that's not needed. What we are doing is `arch_xchg_relaxed` and barrier is not needed.
I did consider adding it to arch/riscv/include/asm/cmpxchg.h but there is limited usage of this primitive and thus kept it limited to usercfi.c
Anyways I'll re-spin removing the barrier.
Consumer of this detects -1 and then return -EFAULT. We would eventually need this when creating shadow stack tokens for kernel shadow stack. I believe `-1` is safe return value which can't be construed as negative kernel address (-EFAULT will be)
Not sure. Didn't see that in x86 or arm64 shadow stack creation. Let me know if its useful.
PROT_READ is pointless here and redundant. I haven't checked if I remove it what happens.
`VM_SHADOW_STACK` takes precedence (take a look at pte_mkwrite and pmd_mkwrite. Only way `VM_SHADOW_STACK` is possible in vmflags is via `map_shadow_stack` or `fork/clone` on existing task with shadow stack enabled.
In a nutshell user can't specify `VM_SHADOW_STACK` directly (indirectly via map_shadow_stack syscall or fork/clone) . But if set in vmaflags then it'll take precedence.
Wouldn't avoiding VM_READ also allow us to get rid of the ugly hack in pte_mkwrite? (VM_WRITE would naturally select the right XWR flags.)
2025-04-23T20:16:58-07:00, Deepak Gupta debug@rivosinc.com:
Ok.
Oof, I'll look why arch is being used like that, thanks.
(I thought arch is only for compiler generated code, so assembly mnemonics would always be defined if the compiler knows them.)
I believe it as well, but I don't see a reason why we need to risk it when we can return the stack value though a pointer and have simple success/failure return value.
It is for automated growing of the stack. I think that the default stack is pointlessly large already, and if other arches don't do it, so we can probably follow their design here as well...
Yeah, I don't like that ugly special case at all, so I was hoping we could somehow avoid it. :)
Userspace specifies CLONE_VM to share address space and spawn new thread. `clone` allow userspace to specify a new stack for new thread. However there is no way to specify new shadow stack base address without changing API. This patch allocates a new shadow stack whenever CLONE_VM is given.
In case of CLONE_VFORK, parent is suspended until child finishes and thus can child use parent shadow stack. In case of !CLONE_VM, COW kicks in because entire address space is copied from parent to child.
`clone3` is extensible and can provide mechanisms using which shadow stack as an input parameter can be provided. This is not settled yet and being extensively discussed on mailing list. Once that's settled, this commit will adapt to that.
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/mmu_context.h | 7 ++ arch/riscv/include/asm/usercfi.h | 25 ++++++++ arch/riscv/kernel/process.c | 9 +++ arch/riscv/kernel/usercfi.c | 120 +++++++++++++++++++++++++++++++++++ 4 files changed, 161 insertions(+)
diff --git a/arch/riscv/include/asm/mmu_context.h b/arch/riscv/include/asm/mmu_context.h index 8c4bc49a3a0f..dbf27a78df6c 100644 --- a/arch/riscv/include/asm/mmu_context.h +++ b/arch/riscv/include/asm/mmu_context.h @@ -48,6 +48,13 @@ static inline unsigned long mm_untag_mask(struct mm_struct *mm) } #endif
+#define deactivate_mm deactivate_mm +static inline void deactivate_mm(struct task_struct *tsk, + struct mm_struct *mm) +{ + shstk_release(tsk); +} + #include <asm-generic/mmu_context.h>
#endif /* _ASM_RISCV_MMU_CONTEXT_H */ diff --git a/arch/riscv/include/asm/usercfi.h b/arch/riscv/include/asm/usercfi.h index 5f2027c51917..82d28ac98d76 100644 --- a/arch/riscv/include/asm/usercfi.h +++ b/arch/riscv/include/asm/usercfi.h @@ -8,6 +8,9 @@ #ifndef __ASSEMBLY__ #include <linux/types.h>
+struct task_struct; +struct kernel_clone_args; + #ifdef CONFIG_RISCV_USER_CFI struct cfi_status { unsigned long ubcfi_en : 1; /* Enable for backward cfi. */ @@ -17,6 +20,28 @@ struct cfi_status { unsigned long shdw_stk_size; /* size of shadow stack */ };
+unsigned long shstk_alloc_thread_stack(struct task_struct *tsk, + const struct kernel_clone_args *args); +void shstk_release(struct task_struct *tsk); +void set_shstk_base(struct task_struct *task, unsigned long shstk_addr, unsigned long size); +unsigned long get_shstk_base(struct task_struct *task, unsigned long *size); +void set_active_shstk(struct task_struct *task, unsigned long shstk_addr); +bool is_shstk_enabled(struct task_struct *task); + +#else + +#define shstk_alloc_thread_stack(tsk, args) 0 + +#define shstk_release(tsk) + +#define get_shstk_base(task, size) 0UL + +#define set_shstk_base(task, shstk_addr, size) + +#define set_active_shstk(task, shstk_addr) + +#define is_shstk_enabled(task) false + #endif /* CONFIG_RISCV_USER_CFI */
#endif /* __ASSEMBLY__ */ diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c index 7c244de77180..99acb6342a37 100644 --- a/arch/riscv/kernel/process.c +++ b/arch/riscv/kernel/process.c @@ -29,6 +29,7 @@ #include <asm/vector.h> #include <asm/cpufeature.h> #include <asm/exec.h> +#include <asm/usercfi.h>
#if defined(CONFIG_STACKPROTECTOR) && !defined(CONFIG_STACKPROTECTOR_PER_TASK) #include <linux/stackprotector.h> @@ -211,6 +212,7 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args) unsigned long clone_flags = args->flags; unsigned long usp = args->stack; unsigned long tls = args->tls; + unsigned long ssp = 0; struct pt_regs *childregs = task_pt_regs(p);
/* Ensure all threads in this mm have the same pointer masking mode. */ @@ -229,11 +231,18 @@ int copy_thread(struct task_struct *p, const struct kernel_clone_args *args) p->thread.s[0] = (unsigned long)args->fn; p->thread.s[1] = (unsigned long)args->fn_arg; } else { + /* allocate new shadow stack if needed. In case of CLONE_VM we have to */ + ssp = shstk_alloc_thread_stack(p, args); + if (IS_ERR_VALUE(ssp)) + return PTR_ERR((void *)ssp); + *childregs = *(current_pt_regs()); /* Turn off status.VS */ riscv_v_vstate_off(childregs); if (usp) /* User fork */ childregs->sp = usp; + /* if needed, set new ssp */ + ssp ? set_active_shstk(p, ssp) : 0; if (clone_flags & CLONE_SETTLS) childregs->tp = tls; childregs->a0 = 0; /* Return value of fork() */ diff --git a/arch/riscv/kernel/usercfi.c b/arch/riscv/kernel/usercfi.c index 24022809a7b5..73cf87dab186 100644 --- a/arch/riscv/kernel/usercfi.c +++ b/arch/riscv/kernel/usercfi.c @@ -19,6 +19,41 @@
#define SHSTK_ENTRY_SIZE sizeof(void *)
+bool is_shstk_enabled(struct task_struct *task) +{ + return task->thread_info.user_cfi_state.ubcfi_en ? true : false; +} + +void set_shstk_base(struct task_struct *task, unsigned long shstk_addr, unsigned long size) +{ + task->thread_info.user_cfi_state.shdw_stk_base = shstk_addr; + task->thread_info.user_cfi_state.shdw_stk_size = size; +} + +unsigned long get_shstk_base(struct task_struct *task, unsigned long *size) +{ + if (size) + *size = task->thread_info.user_cfi_state.shdw_stk_size; + return task->thread_info.user_cfi_state.shdw_stk_base; +} + +void set_active_shstk(struct task_struct *task, unsigned long shstk_addr) +{ + task->thread_info.user_cfi_state.user_shdw_stk = shstk_addr; +} + +/* + * If size is 0, then to be compatible with regular stack we want it to be as big as + * regular stack. Else PAGE_ALIGN it and return back + */ +static unsigned long calc_shstk_size(unsigned long size) +{ + if (size) + return PAGE_ALIGN(size); + + return PAGE_ALIGN(min_t(unsigned long long, rlimit(RLIMIT_STACK), SZ_4G)); +} + /* * Writes on shadow stack can either be `sspush` or `ssamoswap`. `sspush` can happen * implicitly on current shadow stack pointed to by CSR_SSP. `ssamoswap` takes pointer to @@ -142,3 +177,88 @@ SYSCALL_DEFINE3(map_shadow_stack, unsigned long, addr, unsigned long, size, unsi
return allocate_shadow_stack(addr, aligned_size, size, set_tok); } + +/* + * This gets called during clone/clone3/fork. And is needed to allocate a shadow stack for + * cases where CLONE_VM is specified and thus a different stack is specified by user. We + * thus need a separate shadow stack too. How does separate shadow stack is specified by + * user is still being debated. Once that's settled, remove this part of the comment. + * This function simply returns 0 if shadow stack are not supported or if separate shadow + * stack allocation is not needed (like in case of !CLONE_VM) + */ +unsigned long shstk_alloc_thread_stack(struct task_struct *tsk, + const struct kernel_clone_args *args) +{ + unsigned long addr, size; + + /* If shadow stack is not supported, return 0 */ + if (!cpu_supports_shadow_stack()) + return 0; + + /* + * If shadow stack is not enabled on the new thread, skip any + * switch to a new shadow stack. + */ + if (!is_shstk_enabled(tsk)) + return 0; + + /* + * For CLONE_VFORK the child will share the parents shadow stack. + * Set base = 0 and size = 0, this is special means to track this state + * so the freeing logic run for child knows to leave it alone. + */ + if (args->flags & CLONE_VFORK) { + set_shstk_base(tsk, 0, 0); + return 0; + } + + /* + * For !CLONE_VM the child will use a copy of the parents shadow + * stack. + */ + if (!(args->flags & CLONE_VM)) + return 0; + + /* + * reaching here means, CLONE_VM was specified and thus a separate shadow + * stack is needed for new cloned thread. Note: below allocation is happening + * using current mm. + */ + size = calc_shstk_size(args->stack_size); + addr = allocate_shadow_stack(0, size, 0, false); + if (IS_ERR_VALUE(addr)) + return addr; + + set_shstk_base(tsk, addr, size); + + return addr + size; +} + +void shstk_release(struct task_struct *tsk) +{ + unsigned long base = 0, size = 0; + /* If shadow stack is not supported or not enabled, nothing to release */ + if (!cpu_supports_shadow_stack() || !is_shstk_enabled(tsk)) + return; + + /* + * When fork() with CLONE_VM fails, the child (tsk) already has a + * shadow stack allocated, and exit_thread() calls this function to + * free it. In this case the parent (current) and the child share + * the same mm struct. Move forward only when they're same. + */ + if (!tsk->mm || tsk->mm != current->mm) + return; + + /* + * We know shadow stack is enabled but if base is NULL, then + * this task is not managing its own shadow stack (CLONE_VFORK). So + * skip freeing it. + */ + base = get_shstk_base(tsk, &size); + if (!base) + return; + + vm_munmap(base, size); + set_shstk_base(tsk, 0, 0); +}
On 14/03/2025 22:39, Deepak Gupta wrote:
What's the status of this discussion? Can you provide a Link to it?
The use of the ternary here is weird, can you change that to a if statement?
On Tue, Apr 08, 2025 at 12:51:45PM +0200, Alexandre Ghiti wrote:
It's ongoing right now at v15 https://lore.kernel.org/all/20250408-clone3-shadow-stack-v15-0-3fa245c6e3be@...
Implement architecture agnostic prctls() interface for setting and getting shadow stack status.
prctls implemented are PR_GET_SHADOW_STACK_STATUS, PR_SET_SHADOW_STACK_STATUS and PR_LOCK_SHADOW_STACK_STATUS.
As part of PR_SET_SHADOW_STACK_STATUS/PR_GET_SHADOW_STACK_STATUS, only PR_SHADOW_STACK_ENABLE is implemented because RISCV allows each mode to write to their own shadow stack using `sspush` or `ssamoswap`.
PR_LOCK_SHADOW_STACK_STATUS locks current configuration of shadow stack enabling.
Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/usercfi.h | 18 ++++++- arch/riscv/kernel/process.c | 8 +++ arch/riscv/kernel/usercfi.c | 110 +++++++++++++++++++++++++++++++++++++++ 3 files changed, 135 insertions(+), 1 deletion(-)
diff --git a/arch/riscv/include/asm/usercfi.h b/arch/riscv/include/asm/usercfi.h index 82d28ac98d76..c4dcd256f19a 100644 --- a/arch/riscv/include/asm/usercfi.h +++ b/arch/riscv/include/asm/usercfi.h @@ -7,6 +7,7 @@
#ifndef __ASSEMBLY__ #include <linux/types.h> +#include <linux/prctl.h>
struct task_struct; struct kernel_clone_args; @@ -14,7 +15,8 @@ struct kernel_clone_args; #ifdef CONFIG_RISCV_USER_CFI struct cfi_status { unsigned long ubcfi_en : 1; /* Enable for backward cfi. */ - unsigned long rsvd : ((sizeof(unsigned long) * 8) - 1); + unsigned long ubcfi_locked : 1; + unsigned long rsvd : ((sizeof(unsigned long) * 8) - 2); unsigned long user_shdw_stk; /* Current user shadow stack pointer */ unsigned long shdw_stk_base; /* Base address of shadow stack */ unsigned long shdw_stk_size; /* size of shadow stack */ @@ -27,6 +29,12 @@ void set_shstk_base(struct task_struct *task, unsigned long shstk_addr, unsigned unsigned long get_shstk_base(struct task_struct *task, unsigned long *size); void set_active_shstk(struct task_struct *task, unsigned long shstk_addr); bool is_shstk_enabled(struct task_struct *task); +bool is_shstk_locked(struct task_struct *task); +bool is_shstk_allocated(struct task_struct *task); +void set_shstk_lock(struct task_struct *task); +void set_shstk_status(struct task_struct *task, bool enable); + +#define PR_SHADOW_STACK_SUPPORTED_STATUS_MASK (PR_SHADOW_STACK_ENABLE)
#else
@@ -42,6 +50,14 @@ bool is_shstk_enabled(struct task_struct *task);
#define is_shstk_enabled(task) false
+#define is_shstk_locked(task) false + +#define is_shstk_allocated(task) false + +#define set_shstk_lock(task) + +#define set_shstk_status(task, enable) + #endif /* CONFIG_RISCV_USER_CFI */
#endif /* __ASSEMBLY__ */ diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c index 99acb6342a37..cd11667593fe 100644 --- a/arch/riscv/kernel/process.c +++ b/arch/riscv/kernel/process.c @@ -153,6 +153,14 @@ void start_thread(struct pt_regs *regs, unsigned long pc, regs->epc = pc; regs->sp = sp;
+ /* + * clear shadow stack state on exec. + * libc will set it later via prctl. + */ + set_shstk_status(current, false); + set_shstk_base(current, 0, 0); + set_active_shstk(current, 0); + #ifdef CONFIG_64BIT regs->status &= ~SR_UXL;
diff --git a/arch/riscv/kernel/usercfi.c b/arch/riscv/kernel/usercfi.c index 73cf87dab186..b93b324eed26 100644 --- a/arch/riscv/kernel/usercfi.c +++ b/arch/riscv/kernel/usercfi.c @@ -24,6 +24,16 @@ bool is_shstk_enabled(struct task_struct *task) return task->thread_info.user_cfi_state.ubcfi_en ? true : false; }
+bool is_shstk_allocated(struct task_struct *task) +{ + return task->thread_info.user_cfi_state.shdw_stk_base ? true : false; +} + +bool is_shstk_locked(struct task_struct *task) +{ + return task->thread_info.user_cfi_state.ubcfi_locked ? true : false; +} + void set_shstk_base(struct task_struct *task, unsigned long shstk_addr, unsigned long size) { task->thread_info.user_cfi_state.shdw_stk_base = shstk_addr; @@ -42,6 +52,26 @@ void set_active_shstk(struct task_struct *task, unsigned long shstk_addr) task->thread_info.user_cfi_state.user_shdw_stk = shstk_addr; }
+void set_shstk_status(struct task_struct *task, bool enable) +{ + if (!cpu_supports_shadow_stack()) + return; + + task->thread_info.user_cfi_state.ubcfi_en = enable ? 1 : 0; + + if (enable) + task->thread.envcfg |= ENVCFG_SSE; + else + task->thread.envcfg &= ~ENVCFG_SSE; + + csr_write(CSR_ENVCFG, task->thread.envcfg); +} + +void set_shstk_lock(struct task_struct *task) +{ + task->thread_info.user_cfi_state.ubcfi_locked = 1; +} + /* * If size is 0, then to be compatible with regular stack we want it to be as big as * regular stack. Else PAGE_ALIGN it and return back @@ -262,3 +292,83 @@ void shstk_release(struct task_struct *tsk) vm_munmap(base, size); set_shstk_base(tsk, 0, 0); } + +int arch_get_shadow_stack_status(struct task_struct *t, unsigned long __user *status) +{ + unsigned long bcfi_status = 0; + + if (!cpu_supports_shadow_stack()) + return -EINVAL; + + /* this means shadow stack is enabled on the task */ + bcfi_status |= (is_shstk_enabled(t) ? PR_SHADOW_STACK_ENABLE : 0); + + return copy_to_user(status, &bcfi_status, sizeof(bcfi_status)) ? -EFAULT : 0; +} + +int arch_set_shadow_stack_status(struct task_struct *t, unsigned long status) +{ + unsigned long size = 0, addr = 0; + bool enable_shstk = false; + + if (!cpu_supports_shadow_stack()) + return -EINVAL; + + /* Reject unknown flags */ + if (status & ~PR_SHADOW_STACK_SUPPORTED_STATUS_MASK) + return -EINVAL; + + /* bcfi status is locked and further can't be modified by user */ + if (is_shstk_locked(t)) + return -EINVAL; + + enable_shstk = status & PR_SHADOW_STACK_ENABLE; + /* Request is to enable shadow stack and shadow stack is not enabled already */ + if (enable_shstk && !is_shstk_enabled(t)) { + /* shadow stack was allocated and enable request again + * no need to support such usecase and return EINVAL. + */ + if (is_shstk_allocated(t)) + return -EINVAL; + + size = calc_shstk_size(0); + addr = allocate_shadow_stack(0, size, 0, false); + if (IS_ERR_VALUE(addr)) + return -ENOMEM; + set_shstk_base(t, addr, size); + set_active_shstk(t, addr + size); + } + + /* + * If a request to disable shadow stack happens, let's go ahead and release it + * Although, if CLONE_VFORKed child did this, then in that case we will end up + * not releasing the shadow stack (because it might be needed in parent). Although + * we will disable it for VFORKed child. And if VFORKed child tries to enable again + * then in that case, it'll get entirely new shadow stack because following condition + * are true + * - shadow stack was not enabled for vforked child + * - shadow stack base was anyways pointing to 0 + * This shouldn't be a big issue because we want parent to have availability of shadow + * stack whenever VFORKed child releases resources via exit or exec but at the same + * time we want VFORKed child to break away and establish new shadow stack if it desires + * + */ + if (!enable_shstk) + shstk_release(t); + + set_shstk_status(t, enable_shstk); + return 0; +} + +int arch_lock_shadow_stack_status(struct task_struct *task, + unsigned long arg) +{ + /* If shtstk not supported or not enabled on task, nothing to lock here */ + if (!cpu_supports_shadow_stack() || + !is_shstk_enabled(task) || arg != 0) + return -EINVAL; + + set_shstk_lock(task); + + return 0; +}
On Sat, Mar 15, 2025 at 6:51 AM Deepak Gupta debug@rivosinc.com wrote:
LGTM
Reviewed-by: Zong Li zong.li@sifive.com
2025-03-14T14:39:31-07:00, Deepak Gupta debug@rivosinc.com:
The rsvd field shouldn't be necessary as the container for the bitfield is 'unsigned long' sized.
Why don't we use bools here, though? It might produce a better binary and we're not hurting for struct size.
I think that the following is clearer:
return task->thread_info.user_cfi_state.shdw_stk_base
(Similar for all other implicit conversion ternaries.)
There is a new helper we could reuse for this:
envcfg_update_bits(task, ENVCFG_SSE, enable ? ENVCFG_SSE : 0);
Why don't we use the userspace-allocated stack?
I'm completely missing the design idea here... Userspace has absolute over the shadow stack pointer CSR, so we don't need to do much in Linux:
1. interface to set up page tables with -W- PTE and 2. interface to control senvcfg.SSE.
Userspace can do the rest.
The task might want to prevent shadow stack from being enabled?
Thanks.
On Thu, Apr 10, 2025 at 11:45:58AM +0200, Radim Krčmář wrote:
If you remember one of the previous patch discussion, this goes into `thread_info` Don't want to bloat it. Even if we end shoving into task_struct, don't want to bloat that either. I can just convert it into bitmask if bitfields are an eyesore here.
Hmm... noted.
Yeah it's in switch_to.h header. I'll think about it.
Design is like following:
When a user task wants to enable shadow stack for itself, it has to issue a syscall to kernel (like this prctl). Now it can be done independently by user task by first issuing `map_shadow_stack`, then asking kernel to light up envcfg bit and eventually when return to usermode happens, it can write to CSR. It is no different from doing all of the above together in single `prctl` call. They are equivalent in that nature.
Background is that x86 followed this because x86 had workloads/binaries/ functions with (deep)recursive functions and thus by default were forced to always allocate shadow stack to be of the same size as data stack. To reduce burden on userspace for determining and then allocating same size (size of data stack) shadow stack, prctl would do the job of calculating default shadow stack size (and reduce programming error in usermode). arm64 followed the suite. I don't want to find out what's the compatiblity issues we will see and thus just following the suite (given that both approaches are equivalent). Take a look at static `calc_shstk_size(unsigned long size)`.
Coming back to your question of why not allowing userspace to manage its own shadow stack. Answer is that it can manage its own shadow stack. If it does, it just have to be aware of size its allocating for shadow stack.
There is already a patch series going on to manage this using clone3. https://lore.kernel.org/all/20250408-clone3-shadow-stack-v15-4-3fa245c6e3be@...
I fully expect green thread implementations in rust/go or swapcontext based thread management doing this on their own.
Current design is to ensure existing apps dont have to change a lot in userspace and by default kernel gives compatibility. Anyone else wanting to optimize the usage of shadow stack can do so with current design.
-
But Why would it want to do that? Task can simply not issue the prctl. There are glibc tunables as well using which it can be disabled.
Thanks.
2025-04-23T21:44:09-07:00, Deepak Gupta debug@rivosinc.com:
"unsigned long rsvd : ((sizeof(unsigned long) * 8) - 2);"
is an eyesore that defines exactly the same as the two lines alone
unsigned long ubcfi_en : 1; unsigned long ubcfi_locked : 1;
That one should be removed.
If we have only 4 bits in 4/8 bytes, then bitfields do generate worse code than 4 bools and a 0/4 byte hole. The struct size stays the same.
I don't care much about the switch to bools, though, because this code is not called often.
It's just that userspace cannot prevent allocation of the default stack when enabling it, which is the weird part to me. The allocate and enable syscalls could have been nicely composable.
There is already a patch series going on to manage this using clone3. https://lore.kernel.org/all/20250408-clone3-shadow-stack-v15-4-3fa245c6e3be@...
A new ioctl does seem to solve most of the practical issues, thanks.
Right, changing rlimit_stack around shadow stack allocation is not the most elegant way, but it does work.
The task might do it as some last resort to prevent a buggy code from enabling shadow stacks that would just crash. Or whatever complicated reason userspace can think of.
It's more the other way around. I wonder why we're removing this option when we don't really care what userspace does to itself. I think it's complicating the kernel without an obvious gain.
On Thu, Apr 24, 2025 at 03:36:54PM +0200, Radim Krčmář wrote:
I'll remove the bitfields, have single `unsigned long cfi_control_state` And do `#define RISCV_UBCFI_EN 1` and so on.
It just feels wierd. There isn't anything like this for other features lit-up via envcfg. Does hwprobe allow this on per-task basis? I'll look into it.
2025-04-24T11:16:19-07:00, Deepak Gupta debug@rivosinc.com:
I might have seemed too much against the bitfieds, sorry. I am against the rsvd fields, because it is a pointless cognitive overhead and even this series already had a bug in them.
#defines should generate the same code as bitfields (worse than bools), so the source code is really a matter of personal preference. (I do prefer bitfields.)
I think PMM doesn't allow to lock and the rest don't seem configurable from userspace.
It's not that important and we hopefully won't be breaking any userspace if we decided to allow it later, so I'm fine with this version.
Three architectures (x86, aarch64, riscv) have support for indirect branch tracking feature in a very similar fashion. On a very high level, indirect branch tracking is a CPU feature where CPU tracks branches which uses memory operand to perform control transfer in program. As part of this tracking on indirect branches, CPU goes in a state where it expects a landing pad instr on target and if not found then CPU raises some fault (architecture dependent)
x86 landing pad instr - `ENDBRANCH` arch64 landing pad instr - `BTI` riscv landing instr - `lpad`
Given that three major arches have support for indirect branch tracking, This patch makes `prctl` for indirect branch tracking arch agnostic.
To allow userspace to enable this feature for itself, following prtcls are defined: - PR_GET_INDIR_BR_LP_STATUS: Gets current configured status for indirect branch tracking. - PR_SET_INDIR_BR_LP_STATUS: Sets a configuration for indirect branch tracking. Following status options are allowed - PR_INDIR_BR_LP_ENABLE: Enables indirect branch tracking on user thread. - PR_INDIR_BR_LP_DISABLE; Disables indirect branch tracking on user thread. - PR_LOCK_INDIR_BR_LP_STATUS: Locks configured status for indirect branch tracking for user thread.
Signed-off-by: Deepak Gupta debug@rivosinc.com Reviewed-by: Mark Brown broonie@kernel.org --- include/linux/cpu.h | 4 ++++ include/uapi/linux/prctl.h | 27 +++++++++++++++++++++++++++ kernel/sys.c | 30 ++++++++++++++++++++++++++++++ 3 files changed, 61 insertions(+)
diff --git a/include/linux/cpu.h b/include/linux/cpu.h index 6a0a8f1c7c90..fb0c394430c6 100644 --- a/include/linux/cpu.h +++ b/include/linux/cpu.h @@ -204,4 +204,8 @@ static inline bool cpu_mitigations_auto_nosmt(void) } #endif
+int arch_get_indir_br_lp_status(struct task_struct *t, unsigned long __user *status); +int arch_set_indir_br_lp_status(struct task_struct *t, unsigned long status); +int arch_lock_indir_br_lp_status(struct task_struct *t, unsigned long status); + #endif /* _LINUX_CPU_H_ */ diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h index 5c6080680cb2..6cd90460cbad 100644 --- a/include/uapi/linux/prctl.h +++ b/include/uapi/linux/prctl.h @@ -353,4 +353,31 @@ struct prctl_mm_map { */ #define PR_LOCK_SHADOW_STACK_STATUS 76
+/* + * Get the current indirect branch tracking configuration for the current + * thread, this will be the value configured via PR_SET_INDIR_BR_LP_STATUS. + */ +#define PR_GET_INDIR_BR_LP_STATUS 77 + +/* + * Set the indirect branch tracking configuration. PR_INDIR_BR_LP_ENABLE will + * enable cpu feature for user thread, to track all indirect branches and ensure + * they land on arch defined landing pad instruction. + * x86 - If enabled, an indirect branch must land on `ENDBRANCH` instruction. + * arch64 - If enabled, an indirect branch must land on `BTI` instruction. + * riscv - If enabled, an indirect branch must land on `lpad` instruction. + * PR_INDIR_BR_LP_DISABLE will disable feature for user thread and indirect + * branches will no more be tracked by cpu to land on arch defined landing pad + * instruction. + */ +#define PR_SET_INDIR_BR_LP_STATUS 78 +# define PR_INDIR_BR_LP_ENABLE (1UL << 0) + +/* + * Prevent further changes to the specified indirect branch tracking + * configuration. All bits may be locked via this call, including + * undefined bits. + */ +#define PR_LOCK_INDIR_BR_LP_STATUS 79 + #endif /* _LINUX_PRCTL_H */ diff --git a/kernel/sys.c b/kernel/sys.c index cb366ff8703a..f347f3518d0b 100644 --- a/kernel/sys.c +++ b/kernel/sys.c @@ -2336,6 +2336,21 @@ int __weak arch_lock_shadow_stack_status(struct task_struct *t, unsigned long st return -EINVAL; }
+int __weak arch_get_indir_br_lp_status(struct task_struct *t, unsigned long __user *status) +{ + return -EINVAL; +} + +int __weak arch_set_indir_br_lp_status(struct task_struct *t, unsigned long status) +{ + return -EINVAL; +} + +int __weak arch_lock_indir_br_lp_status(struct task_struct *t, unsigned long status) +{ + return -EINVAL; +} + #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
#ifdef CONFIG_ANON_VMA_NAME @@ -2811,6 +2826,21 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, return -EINVAL; error = arch_lock_shadow_stack_status(me, arg2); break; + case PR_GET_INDIR_BR_LP_STATUS: + if (arg3 || arg4 || arg5) + return -EINVAL; + error = arch_get_indir_br_lp_status(me, (unsigned long __user *)arg2); + break; + case PR_SET_INDIR_BR_LP_STATUS: + if (arg3 || arg4 || arg5) + return -EINVAL; + error = arch_set_indir_br_lp_status(me, arg2); + break; + case PR_LOCK_INDIR_BR_LP_STATUS: + if (arg3 || arg4 || arg5) + return -EINVAL; + error = arch_lock_indir_br_lp_status(me, arg2); + break; default: trace_task_prctl_unknown(option, arg2, arg3, arg4, arg5); error = -EINVAL;
On Sat, Mar 15, 2025 at 6:51 AM Deepak Gupta debug@rivosinc.com wrote:
LGTM
Reviewed-by: Zong Li zong.li@sifive.com
On Wed, Apr 09, 2025 at 10:03:05AM +0200, Alexandre Ghiti wrote:
PR_SET_INDIR_BR_LP_STATUS with parameter's bit0 clear will disable branch tracking. This is what arm and riscv settled on for shadow stack enable and disable as well.
prctls implemented are: PR_SET_INDIR_BR_LP_STATUS, PR_GET_INDIR_BR_LP_STATUS and PR_LOCK_INDIR_BR_LP_STATUS
Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/usercfi.h | 16 +++++++- arch/riscv/kernel/entry.S | 2 +- arch/riscv/kernel/process.c | 5 +++ arch/riscv/kernel/usercfi.c | 79 ++++++++++++++++++++++++++++++++++++++++ 4 files changed, 100 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/include/asm/usercfi.h b/arch/riscv/include/asm/usercfi.h index c4dcd256f19a..a8cec7c14d1d 100644 --- a/arch/riscv/include/asm/usercfi.h +++ b/arch/riscv/include/asm/usercfi.h @@ -16,7 +16,9 @@ struct kernel_clone_args; struct cfi_status { unsigned long ubcfi_en : 1; /* Enable for backward cfi. */ unsigned long ubcfi_locked : 1; - unsigned long rsvd : ((sizeof(unsigned long) * 8) - 2); + unsigned long ufcfi_en : 1; /* Enable for forward cfi. Note that ELP goes in sstatus */ + unsigned long ufcfi_locked : 1; + unsigned long rsvd : ((sizeof(unsigned long) * 8) - 4); unsigned long user_shdw_stk; /* Current user shadow stack pointer */ unsigned long shdw_stk_base; /* Base address of shadow stack */ unsigned long shdw_stk_size; /* size of shadow stack */ @@ -33,6 +35,10 @@ bool is_shstk_locked(struct task_struct *task); bool is_shstk_allocated(struct task_struct *task); void set_shstk_lock(struct task_struct *task); void set_shstk_status(struct task_struct *task, bool enable); +bool is_indir_lp_enabled(struct task_struct *task); +bool is_indir_lp_locked(struct task_struct *task); +void set_indir_lp_status(struct task_struct *task, bool enable); +void set_indir_lp_lock(struct task_struct *task);
#define PR_SHADOW_STACK_SUPPORTED_STATUS_MASK (PR_SHADOW_STACK_ENABLE)
@@ -58,6 +64,14 @@ void set_shstk_status(struct task_struct *task, bool enable);
#define set_shstk_status(task, enable)
+#define is_indir_lp_enabled(task) false + +#define is_indir_lp_locked(task) false + +#define set_indir_lp_status(task, enable) + +#define set_indir_lp_lock(task) + #endif /* CONFIG_RISCV_USER_CFI */
#endif /* __ASSEMBLY__ */ diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S index 68c99124ea55..00494b54ff4a 100644 --- a/arch/riscv/kernel/entry.S +++ b/arch/riscv/kernel/entry.S @@ -143,7 +143,7 @@ SYM_CODE_START(handle_exception) * Disable the FPU/Vector to detect illegal usage of floating point * or vector in kernel space. */ - li t0, SR_SUM | SR_FS_VS + li t0, SR_SUM | SR_FS_VS | SR_ELP
REG_L s0, TASK_TI_USER_SP(tp) csrrc s1, CSR_STATUS, t0 diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c index cd11667593fe..4587201dd81d 100644 --- a/arch/riscv/kernel/process.c +++ b/arch/riscv/kernel/process.c @@ -160,6 +160,11 @@ void start_thread(struct pt_regs *regs, unsigned long pc, set_shstk_status(current, false); set_shstk_base(current, 0, 0); set_active_shstk(current, 0); + /* + * disable indirect branch tracking on exec. + * libc will enable it later via prctl. + */ + set_indir_lp_status(current, false);
#ifdef CONFIG_64BIT regs->status &= ~SR_UXL; diff --git a/arch/riscv/kernel/usercfi.c b/arch/riscv/kernel/usercfi.c index b93b324eed26..7937bcef9271 100644 --- a/arch/riscv/kernel/usercfi.c +++ b/arch/riscv/kernel/usercfi.c @@ -72,6 +72,35 @@ void set_shstk_lock(struct task_struct *task) task->thread_info.user_cfi_state.ubcfi_locked = 1; }
+bool is_indir_lp_enabled(struct task_struct *task) +{ + return task->thread_info.user_cfi_state.ufcfi_en ? true : false; +} + +bool is_indir_lp_locked(struct task_struct *task) +{ + return task->thread_info.user_cfi_state.ufcfi_locked ? true : false; +} + +void set_indir_lp_status(struct task_struct *task, bool enable) +{ + if (!cpu_supports_indirect_br_lp_instr()) + return; + + task->thread_info.user_cfi_state.ufcfi_en = enable ? 1 : 0; + + if (enable) + task->thread.envcfg |= ENVCFG_LPE; + else + task->thread.envcfg &= ~ENVCFG_LPE; + + csr_write(CSR_ENVCFG, task->thread.envcfg); +} + +void set_indir_lp_lock(struct task_struct *task) +{ + task->thread_info.user_cfi_state.ufcfi_locked = 1; +} /* * If size is 0, then to be compatible with regular stack we want it to be as big as * regular stack. Else PAGE_ALIGN it and return back @@ -372,3 +401,53 @@ int arch_lock_shadow_stack_status(struct task_struct *task,
return 0; } + +int arch_get_indir_br_lp_status(struct task_struct *t, unsigned long __user *status) +{ + unsigned long fcfi_status = 0; + + if (!cpu_supports_indirect_br_lp_instr()) + return -EINVAL; + + /* indirect branch tracking is enabled on the task or not */ + fcfi_status |= (is_indir_lp_enabled(t) ? PR_INDIR_BR_LP_ENABLE : 0); + + return copy_to_user(status, &fcfi_status, sizeof(fcfi_status)) ? -EFAULT : 0; +} + +int arch_set_indir_br_lp_status(struct task_struct *t, unsigned long status) +{ + bool enable_indir_lp = false; + + if (!cpu_supports_indirect_br_lp_instr()) + return -EINVAL; + + /* indirect branch tracking is locked and further can't be modified by user */ + if (is_indir_lp_locked(t)) + return -EINVAL; + + /* Reject unknown flags */ + if (status & ~PR_INDIR_BR_LP_ENABLE) + return -EINVAL; + + enable_indir_lp = (status & PR_INDIR_BR_LP_ENABLE) ? true : false; + set_indir_lp_status(t, enable_indir_lp); + + return 0; +} + +int arch_lock_indir_br_lp_status(struct task_struct *task, + unsigned long arg) +{ + /* + * If indirect branch tracking is not supported or not enabled on task, + * nothing to lock here + */ + if (!cpu_supports_indirect_br_lp_instr() || + !is_indir_lp_enabled(task) || arg != 0) + return -EINVAL; + + set_indir_lp_lock(task); + + return 0; +}
On Sat, Mar 15, 2025 at 6:51 AM Deepak Gupta debug@rivosinc.com wrote:
LGTM
Reviewed-by: Zong Li zong.li@sifive.com
zicfiss / zicfilp introduces a new exception to priv isa `software check exception` with cause code = 18. This patch implements software check exception.
Additionally it implements a cfi violation handler which checks for code in xtval. If xtval=2, it means that sw check exception happened because of an indirect branch not landing on 4 byte aligned PC or not landing on `lpad` instruction or label value embedded in `lpad` not matching label value setup in `x7`. If xtval=3, it means that sw check exception happened because of mismatch between link register (x1 or x5) and top of shadow stack (on execution of `sspopchk`).
In case of cfi violation, SIGSEGV is raised with code=SEGV_CPERR. SEGV_CPERR was introduced by x86 shadow stack patches.
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/asm-prototypes.h | 1 + arch/riscv/include/asm/entry-common.h | 2 ++ arch/riscv/kernel/entry.S | 3 +++ arch/riscv/kernel/traps.c | 43 +++++++++++++++++++++++++++++++++ 4 files changed, 49 insertions(+)
diff --git a/arch/riscv/include/asm/asm-prototypes.h b/arch/riscv/include/asm/asm-prototypes.h index cd627ec289f1..5a27cefd7805 100644 --- a/arch/riscv/include/asm/asm-prototypes.h +++ b/arch/riscv/include/asm/asm-prototypes.h @@ -51,6 +51,7 @@ DECLARE_DO_ERROR_INFO(do_trap_ecall_u); DECLARE_DO_ERROR_INFO(do_trap_ecall_s); DECLARE_DO_ERROR_INFO(do_trap_ecall_m); DECLARE_DO_ERROR_INFO(do_trap_break); +DECLARE_DO_ERROR_INFO(do_trap_software_check);
asmlinkage void handle_bad_stack(struct pt_regs *regs); asmlinkage void do_page_fault(struct pt_regs *regs); diff --git a/arch/riscv/include/asm/entry-common.h b/arch/riscv/include/asm/entry-common.h index b28ccc6cdeea..34ed149af5d1 100644 --- a/arch/riscv/include/asm/entry-common.h +++ b/arch/riscv/include/asm/entry-common.h @@ -40,4 +40,6 @@ static inline int handle_misaligned_store(struct pt_regs *regs) } #endif
+bool handle_user_cfi_violation(struct pt_regs *regs); + #endif /* _ASM_RISCV_ENTRY_COMMON_H */ diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S index 00494b54ff4a..9c00cac3f6f2 100644 --- a/arch/riscv/kernel/entry.S +++ b/arch/riscv/kernel/entry.S @@ -472,6 +472,9 @@ SYM_DATA_START_LOCAL(excp_vect_table) RISCV_PTR do_page_fault /* load page fault */ RISCV_PTR do_trap_unknown RISCV_PTR do_page_fault /* store page fault */ + RISCV_PTR do_trap_unknown /* cause=16 */ + RISCV_PTR do_trap_unknown /* cause=17 */ + RISCV_PTR do_trap_software_check /* cause=18 is sw check exception */ SYM_DATA_END_LABEL(excp_vect_table, SYM_L_LOCAL, excp_vect_table_end)
#ifndef CONFIG_MMU diff --git a/arch/riscv/kernel/traps.c b/arch/riscv/kernel/traps.c index 8ff8e8b36524..3f7709f4595a 100644 --- a/arch/riscv/kernel/traps.c +++ b/arch/riscv/kernel/traps.c @@ -354,6 +354,49 @@ void do_trap_ecall_u(struct pt_regs *regs)
}
+#define CFI_TVAL_FCFI_CODE 2 +#define CFI_TVAL_BCFI_CODE 3 +/* handle cfi violations */ +bool handle_user_cfi_violation(struct pt_regs *regs) +{ + bool ret = false; + unsigned long tval = csr_read(CSR_TVAL); + + if ((tval == CFI_TVAL_FCFI_CODE && cpu_supports_indirect_br_lp_instr()) || + (tval == CFI_TVAL_BCFI_CODE && cpu_supports_shadow_stack())) { + do_trap_error(regs, SIGSEGV, SEGV_CPERR, regs->epc, + "Oops - control flow violation"); + ret = true; + } + + return ret; +} + +/* + * software check exception is defined with risc-v cfi spec. Software check + * exception is raised when:- + * a) An indirect branch doesn't land on 4 byte aligned PC or `lpad` + * instruction or `label` value programmed in `lpad` instr doesn't + * match with value setup in `x7`. reported code in `xtval` is 2. + * b) `sspopchk` instruction finds a mismatch between top of shadow stack (ssp) + * and x1/x5. reported code in `xtval` is 3. + */ +asmlinkage __visible __trap_section void do_trap_software_check(struct pt_regs *regs) +{ + if (user_mode(regs)) { + irqentry_enter_from_user_mode(regs); + + /* not a cfi violation, then merge into flow of unknown trap handler */ + if (!handle_user_cfi_violation(regs)) + do_trap_unknown(regs); + + irqentry_exit_to_user_mode(regs); + } else { + /* sw check exception coming from kernel is a bug in kernel */ + die(regs, "Kernel BUG"); + } +} + #ifdef CONFIG_MMU asmlinkage __visible noinstr void do_page_fault(struct pt_regs *regs) {
From: Andy Chiu andybnac@gmail.com
The function save_v_state() served two purposes. First, it saved extension context into the signal stack. Then, it constructed the extension header if there was no fault. The second part is independent of the extension itself. As a result, we can pull that part out, so future extensions may reuse it. This patch adds arch_ext_list and makes setup_sigcontext() go through all possible extensions' save() callback. The callback returns a positive value indicating the size of the successfully saved extension. Then the kernel proceeds to construct the header for that extension. The kernel skips an extension if it does not exist, or if the saving fails for some reasons. The error code is propagated out on the later case.
This patch does not introduce any functional changes.
Signed-off-by: Andy Chiu andybnac@gmail.com --- arch/riscv/include/asm/vector.h | 3 ++ arch/riscv/kernel/signal.c | 62 +++++++++++++++++++++++++++-------------- 2 files changed, 44 insertions(+), 21 deletions(-)
diff --git a/arch/riscv/include/asm/vector.h b/arch/riscv/include/asm/vector.h index e8a83f55be2b..05390538ea8a 100644 --- a/arch/riscv/include/asm/vector.h +++ b/arch/riscv/include/asm/vector.h @@ -407,6 +407,9 @@ static inline bool riscv_v_vstate_ctrl_user_allowed(void) { return false; } #define riscv_v_thread_free(tsk) do {} while (0) #define riscv_v_setup_ctx_cache() do {} while (0) #define riscv_v_thread_alloc(tsk) do {} while (0) +#define get_cpu_vector_context() do {} while (0) +#define put_cpu_vector_context() do {} while (0) +#define riscv_v_vstate_set_restore(task, regs) do {} while (0)
#endif /* CONFIG_RISCV_ISA_V */
diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c index 94e905eea1de..80c70dccf09f 100644 --- a/arch/riscv/kernel/signal.c +++ b/arch/riscv/kernel/signal.c @@ -68,18 +68,19 @@ static long save_fp_state(struct pt_regs *regs, #define restore_fp_state(task, regs) (0) #endif
-#ifdef CONFIG_RISCV_ISA_V - -static long save_v_state(struct pt_regs *regs, void __user **sc_vec) +static long save_v_state(struct pt_regs *regs, void __user *sc_vec) { - struct __riscv_ctx_hdr __user *hdr; struct __sc_riscv_v_state __user *state; void __user *datap; long err;
- hdr = *sc_vec; - /* Place state to the user's signal context space after the hdr */ - state = (struct __sc_riscv_v_state __user *)(hdr + 1); + if (!IS_ENABLED(CONFIG_RISCV_ISA_V) || + !((has_vector() || has_xtheadvector()) && + riscv_v_vstate_query(regs))) + return 0; + + /* Place state to the user's signal context spac */ + state = (struct __sc_riscv_v_state __user *)sc_vec; /* Point datap right after the end of __sc_riscv_v_state */ datap = state + 1;
@@ -97,15 +98,11 @@ static long save_v_state(struct pt_regs *regs, void __user **sc_vec) err |= __put_user((__force void *)datap, &state->v_state.datap); /* Copy the whole vector content to user space datap. */ err |= __copy_to_user(datap, current->thread.vstate.datap, riscv_v_vsize); - /* Copy magic to the user space after saving all vector conetext */ - err |= __put_user(RISCV_V_MAGIC, &hdr->magic); - err |= __put_user(riscv_v_sc_size, &hdr->size); if (unlikely(err)) - return err; + return -EFAULT;
- /* Only progress the sv_vec if everything has done successfully */ - *sc_vec += riscv_v_sc_size; - return 0; + /* Only return the size if everything has done successfully */ + return riscv_v_sc_size; }
/* @@ -142,10 +139,20 @@ static long __restore_v_state(struct pt_regs *regs, void __user *sc_vec) */ return copy_from_user(current->thread.vstate.datap, datap, riscv_v_vsize); } -#else -#define save_v_state(task, regs) (0) -#define __restore_v_state(task, regs) (0) -#endif + +struct arch_ext_priv { + __u32 magic; + long (*save)(struct pt_regs *regs, void __user *sc_vec); +}; + +struct arch_ext_priv arch_ext_list[] = { + { + .magic = RISCV_V_MAGIC, + .save = &save_v_state, + }, +}; + +const size_t nr_arch_exts = ARRAY_SIZE(arch_ext_list);
static long restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc) @@ -276,7 +283,8 @@ static long setup_sigcontext(struct rt_sigframe __user *frame, { struct sigcontext __user *sc = &frame->uc.uc_mcontext; struct __riscv_ctx_hdr __user *sc_ext_ptr = &sc->sc_extdesc.hdr; - long err; + struct arch_ext_priv *arch_ext; + long err, i, ext_size;
/* sc_regs is structured the same as the start of pt_regs */ err = __copy_to_user(&sc->sc_regs, regs, sizeof(sc->sc_regs)); @@ -284,8 +292,20 @@ static long setup_sigcontext(struct rt_sigframe __user *frame, if (has_fpu()) err |= save_fp_state(regs, &sc->sc_fpregs); /* Save the vector state. */ - if ((has_vector() || has_xtheadvector()) && riscv_v_vstate_query(regs)) - err |= save_v_state(regs, (void __user **)&sc_ext_ptr); + for (i = 0; i < nr_arch_exts; i++) { + arch_ext = &arch_ext_list[i]; + if (!arch_ext->save) + continue; + + ext_size = arch_ext->save(regs, sc_ext_ptr + 1); + if (ext_size <= 0) { + err |= ext_size; + } else { + err |= __put_user(arch_ext->magic, &sc_ext_ptr->magic); + err |= __put_user(ext_size, &sc_ext_ptr->size); + sc_ext_ptr = (void *)sc_ext_ptr + ext_size; + } + } /* Write zero to fp-reserved space and check it on restore_sigcontext */ err |= __put_user(0, &sc->sc_extdesc.reserved); /* And put END __riscv_ctx_hdr at the end. */
Save shadow stack pointer in sigcontext structure while delivering signal. Restore shadow stack pointer from sigcontext on sigreturn.
As part of save operation, kernel uses `ssamoswap` to save snapshot of current shadow stack on shadow stack itself (can be called as a save token). During restore on sigreturn, kernel retrieves token from top of shadow stack and validates it. This allows that user mode can't arbitrary pivot to any shadow stack address without having a token and thus provide strong security assurance between signaly delivery and sigreturn window.
Use ABI compatible way of saving/restoring shadow stack pointer into signal stack. This follows what Vector extension, where extra registers are placed in a form of extension header + extension body in the stack. The extension header indicates the size of the extra architectural states plus the size of header itself, and a magic identifier of the extension. Then, the extensions body contains the new architectural states in the form defined by uapi.
Signed-off-by: Andy Chiu andy.chiu@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/asm/usercfi.h | 10 ++++ arch/riscv/include/uapi/asm/ptrace.h | 4 ++ arch/riscv/include/uapi/asm/sigcontext.h | 1 + arch/riscv/kernel/signal.c | 80 ++++++++++++++++++++++++++++++++ arch/riscv/kernel/usercfi.c | 56 ++++++++++++++++++++++ 5 files changed, 151 insertions(+)
diff --git a/arch/riscv/include/asm/usercfi.h b/arch/riscv/include/asm/usercfi.h index a8cec7c14d1d..361f59edbdef 100644 --- a/arch/riscv/include/asm/usercfi.h +++ b/arch/riscv/include/asm/usercfi.h @@ -8,6 +8,7 @@ #ifndef __ASSEMBLY__ #include <linux/types.h> #include <linux/prctl.h> +#include <linux/errno.h>
struct task_struct; struct kernel_clone_args; @@ -35,6 +36,9 @@ bool is_shstk_locked(struct task_struct *task); bool is_shstk_allocated(struct task_struct *task); void set_shstk_lock(struct task_struct *task); void set_shstk_status(struct task_struct *task, bool enable); +unsigned long get_active_shstk(struct task_struct *task); +int restore_user_shstk(struct task_struct *tsk, unsigned long shstk_ptr); +int save_user_shstk(struct task_struct *tsk, unsigned long *saved_shstk_ptr); bool is_indir_lp_enabled(struct task_struct *task); bool is_indir_lp_locked(struct task_struct *task); void set_indir_lp_status(struct task_struct *task, bool enable); @@ -72,6 +76,12 @@ void set_indir_lp_lock(struct task_struct *task);
#define set_indir_lp_lock(task)
+#define restore_user_shstk(tsk, shstk_ptr) -EINVAL + +#define save_user_shstk(tsk, saved_shstk_ptr) -EINVAL + +#define get_active_shstk(task) 0UL + #endif /* CONFIG_RISCV_USER_CFI */
#endif /* __ASSEMBLY__ */ diff --git a/arch/riscv/include/uapi/asm/ptrace.h b/arch/riscv/include/uapi/asm/ptrace.h index a38268b19c3d..659ea3af5680 100644 --- a/arch/riscv/include/uapi/asm/ptrace.h +++ b/arch/riscv/include/uapi/asm/ptrace.h @@ -127,6 +127,10 @@ struct __riscv_v_regset_state { */ #define RISCV_MAX_VLENB (8192)
+struct __sc_riscv_cfi_state { + unsigned long ss_ptr; /* shadow stack pointer */ +}; + #endif /* __ASSEMBLY__ */
#endif /* _UAPI_ASM_RISCV_PTRACE_H */ diff --git a/arch/riscv/include/uapi/asm/sigcontext.h b/arch/riscv/include/uapi/asm/sigcontext.h index cd4f175dc837..f37e4beffe03 100644 --- a/arch/riscv/include/uapi/asm/sigcontext.h +++ b/arch/riscv/include/uapi/asm/sigcontext.h @@ -10,6 +10,7 @@
/* The Magic number for signal context frame header. */ #define RISCV_V_MAGIC 0x53465457 +#define RISCV_ZICFISS_MAGIC 0x9487 #define END_MAGIC 0x0
/* The size of END signal context header. */ diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c index 80c70dccf09f..a7472a6fcdca 100644 --- a/arch/riscv/kernel/signal.c +++ b/arch/riscv/kernel/signal.c @@ -22,11 +22,13 @@ #include <asm/vector.h> #include <asm/csr.h> #include <asm/cacheflush.h> +#include <asm/usercfi.h>
unsigned long signal_minsigstksz __ro_after_init;
extern u32 __user_rt_sigreturn[2]; static size_t riscv_v_sc_size __ro_after_init; +static size_t riscv_zicfiss_sc_size __ro_after_init;
#define DEBUG_SIG 0
@@ -140,6 +142,62 @@ static long __restore_v_state(struct pt_regs *regs, void __user *sc_vec) return copy_from_user(current->thread.vstate.datap, datap, riscv_v_vsize); }
+static long save_cfiss_state(struct pt_regs *regs, void __user *sc_cfi) +{ + struct __sc_riscv_cfi_state __user *state = sc_cfi; + unsigned long ss_ptr = 0; + long err = 0; + + if (!IS_ENABLED(CONFIG_RISCV_USER_CFI) || !is_shstk_enabled(current)) + return 0; + + /* + * Save a pointer to shadow stack itself on shadow stack as a form of token. + * A token on shadow gives following properties + * - Safe save and restore for shadow stack switching. Any save of shadow stack + * must have had saved a token on shadow stack. Similarly any restore of shadow + * stack must check the token before restore. Since writing to shadow stack with + * address of shadow stack itself is not easily allowed. A restore without a save + * is quite difficult for an attacker to perform. + * - A natural break. A token in shadow stack provides a natural break in shadow stack + * So a single linear range can be bucketed into different shadow stack segments. Any + * sspopchk will detect the condition and fault to kernel as sw check exception. + */ + err |= save_user_shstk(current, &ss_ptr); + err |= __put_user(ss_ptr, &state->ss_ptr); + if (unlikely(err)) + return -EFAULT; + + return riscv_zicfiss_sc_size; +} + +static long __restore_cfiss_state(struct pt_regs *regs, void __user *sc_cfi) +{ + struct __sc_riscv_cfi_state __user *state = sc_cfi; + unsigned long ss_ptr = 0; + long err; + + /* + * Restore shadow stack as a form of token stored on shadow stack itself as a safe + * way to restore. + * A token on shadow gives following properties + * - Safe save and restore for shadow stack switching. Any save of shadow stack + * must have had saved a token on shadow stack. Similarly any restore of shadow + * stack must check the token before restore. Since writing to shadow stack with + * address of shadow stack itself is not easily allowed. A restore without a save + * is quite difficult for an attacker to perform. + * - A natural break. A token in shadow stack provides a natural break in shadow stack + * So a single linear range can be bucketed into different shadow stack segments. + * sspopchk will detect the condition and fault to kernel as sw check exception. + */ + err = __copy_from_user(&ss_ptr, &state->ss_ptr, sizeof(unsigned long)); + + if (unlikely(err)) + return err; + + return restore_user_shstk(current, ss_ptr); +} + struct arch_ext_priv { __u32 magic; long (*save)(struct pt_regs *regs, void __user *sc_vec); @@ -150,6 +208,10 @@ struct arch_ext_priv arch_ext_list[] = { .magic = RISCV_V_MAGIC, .save = &save_v_state, }, + { + .magic = RISCV_ZICFISS_MAGIC, + .save = &save_cfiss_state, + }, };
const size_t nr_arch_exts = ARRAY_SIZE(arch_ext_list); @@ -202,6 +264,12 @@ static long restore_sigcontext(struct pt_regs *regs,
err = __restore_v_state(regs, sc_ext_ptr); break; + case RISCV_ZICFISS_MAGIC: + if (!is_shstk_enabled(current) || size != riscv_zicfiss_sc_size) + return -EINVAL; + + err = __restore_cfiss_state(regs, sc_ext_ptr); + break; default: return -EINVAL; } @@ -222,6 +290,10 @@ static size_t get_rt_frame_size(bool cal_all) if (cal_all || riscv_v_vstate_query(task_pt_regs(current))) total_context_size += riscv_v_sc_size; } + + if (is_shstk_enabled(current)) + total_context_size += riscv_zicfiss_sc_size; + /* * Preserved a __riscv_ctx_hdr for END signal context header if an * extension uses __riscv_extra_ext_header @@ -365,6 +437,11 @@ static int setup_rt_frame(struct ksignal *ksig, sigset_t *set, #ifdef CONFIG_MMU regs->ra = (unsigned long)VDSO_SYMBOL( current->mm->context.vdso, rt_sigreturn); + + /* if bcfi is enabled x1 (ra) and x5 (t0) must match. not sure if we need this? */ + if (is_shstk_enabled(current)) + regs->t0 = regs->ra; + #else /* * For the nommu case we don't have a VDSO. Instead we push two @@ -493,6 +570,9 @@ void __init init_rt_signal_env(void) { riscv_v_sc_size = sizeof(struct __riscv_ctx_hdr) + sizeof(struct __sc_riscv_v_state) + riscv_v_vsize; + + riscv_zicfiss_sc_size = sizeof(struct __riscv_ctx_hdr) + + sizeof(struct __sc_riscv_cfi_state); /* * Determine the stack space required for guaranteed signal delivery. * The signal_minsigstksz will be populated into the AT_MINSIGSTKSZ entry diff --git a/arch/riscv/kernel/usercfi.c b/arch/riscv/kernel/usercfi.c index 7937bcef9271..d31d89618763 100644 --- a/arch/riscv/kernel/usercfi.c +++ b/arch/riscv/kernel/usercfi.c @@ -52,6 +52,11 @@ void set_active_shstk(struct task_struct *task, unsigned long shstk_addr) task->thread_info.user_cfi_state.user_shdw_stk = shstk_addr; }
+unsigned long get_active_shstk(struct task_struct *task) +{ + return task->thread_info.user_cfi_state.user_shdw_stk; +} + void set_shstk_status(struct task_struct *task, bool enable) { if (!cpu_supports_shadow_stack()) @@ -170,6 +175,57 @@ static int create_rstor_token(unsigned long ssp, unsigned long *token_addr) return 0; }
+/* + * Save user shadow stack pointer on shadow stack itself and return pointer to saved location + * returns -EFAULT if operation was unsuccessful + */ +int save_user_shstk(struct task_struct *tsk, unsigned long *saved_shstk_ptr) +{ + unsigned long ss_ptr = 0; + unsigned long token_loc = 0; + int ret = 0; + + if (saved_shstk_ptr == NULL) + return -EINVAL; + + ss_ptr = get_active_shstk(tsk); + ret = create_rstor_token(ss_ptr, &token_loc); + + if (!ret) { + *saved_shstk_ptr = token_loc; + set_active_shstk(tsk, token_loc); + } + + return ret; +} + +/* + * Restores user shadow stack pointer from token on shadow stack for task `tsk` + * returns -EFAULT if operation was unsuccessful + */ +int restore_user_shstk(struct task_struct *tsk, unsigned long shstk_ptr) +{ + unsigned long token = 0; + + token = amo_user_shstk((unsigned long __user *)shstk_ptr, 0); + + if (token == -1) + return -EFAULT; + + /* invalid token, return EINVAL */ + if ((token - shstk_ptr) != SHSTK_ENTRY_SIZE) { + pr_info_ratelimited( + "%s[%d]: bad restore token in %s: pc=%p sp=%p, token=%p, shstk_ptr=%p\n", + tsk->comm, task_pid_nr(tsk), __func__, (void *)(task_pt_regs(tsk)->epc), + (void *)(task_pt_regs(tsk)->sp), (void *)token, (void *)shstk_ptr); + return -EINVAL; + } + + /* all checks passed, set active shstk and return success */ + set_active_shstk(tsk, token); + return 0; +} + static unsigned long allocate_shadow_stack(unsigned long addr, unsigned long size, unsigned long token_offset, bool set_tok) {
2025-03-14T14:39:36-07:00, Deepak Gupta debug@rivosinc.com:
!is_shstk_enabled() should be enough here.
[The rest looks fine, but I'll need more time to think about it. I'll return to this patch in v13.]
Updating __show_regs to print captured shadow stack pointer as well. On tasks where shadow stack is disabled, it'll simply print 0.
Signed-off-by: Deepak Gupta debug@rivosinc.com Reviewed-by: Alexandre Ghiti alexghiti@rivosinc.com --- arch/riscv/kernel/process.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c index 4587201dd81d..6bb53ce72ed5 100644 --- a/arch/riscv/kernel/process.c +++ b/arch/riscv/kernel/process.c @@ -90,8 +90,8 @@ void __show_regs(struct pt_regs *regs) regs->s8, regs->s9, regs->s10); pr_cont(" s11: " REG_FMT " t3 : " REG_FMT " t4 : " REG_FMT "\n", regs->s11, regs->t3, regs->t4); - pr_cont(" t5 : " REG_FMT " t6 : " REG_FMT "\n", - regs->t5, regs->t6); + pr_cont(" t5 : " REG_FMT " t6 : " REG_FMT " ssp : " REG_FMT "\n", + regs->t5, regs->t6, get_active_shstk(current));
pr_cont("status: " REG_FMT " badaddr: " REG_FMT " cause: " REG_FMT "\n", regs->status, regs->badaddr, regs->cause);
Expose a new register type NT_RISCV_USER_CFI for risc-v cfi status and state. Intentionally both landing pad and shadow stack status and state are rolled into cfi state. Creating two different NT_RISCV_USER_XXX would not be useful and wastage of a note type. Enabling or disabling of feature is not allowed via ptrace set interface. However setting `elp` state or setting shadow stack pointer are allowed via ptrace set interface. It is expected `gdb` might have use to fixup `elp` state or `shadow stack` pointer.
Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/uapi/asm/ptrace.h | 18 ++++++++ arch/riscv/kernel/ptrace.c | 83 ++++++++++++++++++++++++++++++++++++ include/uapi/linux/elf.h | 1 + 3 files changed, 102 insertions(+)
diff --git a/arch/riscv/include/uapi/asm/ptrace.h b/arch/riscv/include/uapi/asm/ptrace.h index 659ea3af5680..e6571fba8a8a 100644 --- a/arch/riscv/include/uapi/asm/ptrace.h +++ b/arch/riscv/include/uapi/asm/ptrace.h @@ -131,6 +131,24 @@ struct __sc_riscv_cfi_state { unsigned long ss_ptr; /* shadow stack pointer */ };
+struct __cfi_status { + /* indirect branch tracking state */ + __u64 lp_en : 1; + __u64 lp_lock : 1; + __u64 elp_state : 1; + + /* shadow stack status */ + __u64 shstk_en : 1; + __u64 shstk_lock : 1; + + __u64 rsvd : sizeof(__u64) - 5; +}; + +struct user_cfi_state { + struct __cfi_status cfi_status; + __u64 shstk_ptr; +}; + #endif /* __ASSEMBLY__ */
#endif /* _UAPI_ASM_RISCV_PTRACE_H */ diff --git a/arch/riscv/kernel/ptrace.c b/arch/riscv/kernel/ptrace.c index ea67e9fb7a58..df8b7c6ab671 100644 --- a/arch/riscv/kernel/ptrace.c +++ b/arch/riscv/kernel/ptrace.c @@ -19,6 +19,7 @@ #include <linux/regset.h> #include <linux/sched.h> #include <linux/sched/task_stack.h> +#include <asm/usercfi.h>
enum riscv_regset { REGSET_X, @@ -31,6 +32,9 @@ enum riscv_regset { #ifdef CONFIG_RISCV_ISA_SUPM REGSET_TAGGED_ADDR_CTRL, #endif +#ifdef CONFIG_RISCV_USER_CFI + REGSET_CFI, +#endif };
static int riscv_gpr_get(struct task_struct *target, @@ -184,6 +188,75 @@ static int tagged_addr_ctrl_set(struct task_struct *target, } #endif
+#ifdef CONFIG_RISCV_USER_CFI +static int riscv_cfi_get(struct task_struct *target, + const struct user_regset *regset, + struct membuf to) +{ + struct user_cfi_state user_cfi; + struct pt_regs *regs; + + regs = task_pt_regs(target); + + user_cfi.cfi_status.lp_en = is_indir_lp_enabled(target); + user_cfi.cfi_status.lp_lock = is_indir_lp_locked(target); + user_cfi.cfi_status.elp_state = (regs->status & SR_ELP); + + user_cfi.cfi_status.shstk_en = is_shstk_enabled(target); + user_cfi.cfi_status.shstk_lock = is_shstk_locked(target); + user_cfi.shstk_ptr = get_active_shstk(target); + + return membuf_write(&to, &user_cfi, sizeof(user_cfi)); +} + +/* + * Does it make sense to allowing enable / disable of cfi via ptrace? + * Not allowing enable / disable / locking control via ptrace for now. + * Setting shadow stack pointer is allowed. GDB might use it to unwind or + * some other fixup. Similarly gdb might want to suppress elp and may want + * to reset elp state. + */ +static int riscv_cfi_set(struct task_struct *target, + const struct user_regset *regset, + unsigned int pos, unsigned int count, + const void *kbuf, const void __user *ubuf) +{ + int ret; + struct user_cfi_state user_cfi; + struct pt_regs *regs; + + regs = task_pt_regs(target); + + ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, &user_cfi, 0, -1); + if (ret) + return ret; + + /* + * Not allowing enabling or locking shadow stack or landing pad + * There is no disabling of shadow stack or landing pad via ptrace + * rsvd field should be set to zero so that if those fields are needed in future + */ + if (user_cfi.cfi_status.lp_en || user_cfi.cfi_status.lp_lock || + user_cfi.cfi_status.shstk_en || user_cfi.cfi_status.shstk_lock || + !user_cfi.cfi_status.rsvd) + return -EINVAL; + + /* If lpad is enabled on target and ptrace requests to set / clear elp, do that */ + if (is_indir_lp_enabled(target)) { + if (user_cfi.cfi_status.elp_state) /* set elp state */ + regs->status |= SR_ELP; + else + regs->status &= ~SR_ELP; /* clear elp state */ + } + + /* If shadow stack enabled on target, set new shadow stack pointer */ + if (is_shstk_enabled(target)) + set_active_shstk(target, user_cfi.shstk_ptr); + + return 0; +} +#endif + static const struct user_regset riscv_user_regset[] = { [REGSET_X] = { .core_note_type = NT_PRSTATUS, @@ -224,6 +297,16 @@ static const struct user_regset riscv_user_regset[] = { .set = tagged_addr_ctrl_set, }, #endif +#ifdef CONFIG_RISCV_USER_CFI + [REGSET_CFI] = { + .core_note_type = NT_RISCV_USER_CFI, + .align = sizeof(__u64), + .n = sizeof(struct user_cfi_state) / sizeof(__u64), + .size = sizeof(__u64), + .regset_get = riscv_cfi_get, + .set = riscv_cfi_set, + }, +#endif };
static const struct user_regset_view riscv_user_native_view = { diff --git a/include/uapi/linux/elf.h b/include/uapi/linux/elf.h index b44069d29cec..b9daed4ab780 100644 --- a/include/uapi/linux/elf.h +++ b/include/uapi/linux/elf.h @@ -452,6 +452,7 @@ typedef struct elf64_shdr { #define NT_RISCV_CSR 0x900 /* RISC-V Control and Status Registers */ #define NT_RISCV_VECTOR 0x901 /* RISC-V vector registers */ #define NT_RISCV_TAGGED_ADDR_CTRL 0x902 /* RISC-V tagged address control (prctl()) */ +#define NT_RISCV_USER_CFI 0x903 /* RISC-V shadow stack state */ #define NT_LOONGARCH_CPUCFG 0xa00 /* LoongArch CPU config registers */ #define NT_LOONGARCH_CSR 0xa01 /* LoongArch control and status registers */ #define NT_LOONGARCH_LSX 0xa02 /* LoongArch Loongson SIMD Extension registers */
2025-03-14T14:39:38-07:00, Deepak Gupta debug@rivosinc.com:
I remember there was deep hatred towards bitfields in the Linux community, have things changes?
I think you meant "64 - 5".
Why not `size = sizeof(struct user_cfi_state)` and `n = 1`?
[I haven't yet reviewed if a new register is the right thing to do nor looked at the rest of the patch.]
On Thu, Mar 20, 2025 at 3:24 PM Radim Krčmář rkrcmar@ventanamicro.com wrote:
hmm. I didn't know about the strong hatred. Although I can see lots of examples of this pattern in existing kernel code. No strong feelings on my side, I can change this and have it single 64bit field and accessed via bitmasks.
eeh. bad bug. thanks.
yeah another good catch. Should write a kselftest against it, so that it can be caught.
2025-03-20T16:09:12-07:00, Deepak Gupta debug@rivosinc.com:
There is a good reason for it. :)
The C standard left important behavior as implementation-specific (by mistake, I hope). I do like bitfields, but you have to be extra careful when working with them.
This is uapi and bitfields do not specify the internal representation. A program compiled at a different time can see completely different order of the bitfields, so the uapi would break.
We cannot use bitfields here.
Adding enumeration of zicfilp and zicfiss extensions in hwprobe syscall.
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/include/uapi/asm/hwprobe.h | 2 ++ arch/riscv/kernel/sys_hwprobe.c | 2 ++ 2 files changed, 4 insertions(+)
diff --git a/arch/riscv/include/uapi/asm/hwprobe.h b/arch/riscv/include/uapi/asm/hwprobe.h index c3c1cc951cb9..c1b537b50158 100644 --- a/arch/riscv/include/uapi/asm/hwprobe.h +++ b/arch/riscv/include/uapi/asm/hwprobe.h @@ -73,6 +73,8 @@ struct riscv_hwprobe { #define RISCV_HWPROBE_EXT_ZCMOP (1ULL << 47) #define RISCV_HWPROBE_EXT_ZAWRS (1ULL << 48) #define RISCV_HWPROBE_EXT_SUPM (1ULL << 49) +#define RISCV_HWPROBE_EXT_ZICFILP (1ULL << 50) +#define RISCV_HWPROBE_EXT_ZICFISS (1ULL << 51) #define RISCV_HWPROBE_KEY_CPUPERF_0 5 #define RISCV_HWPROBE_MISALIGNED_UNKNOWN (0 << 0) #define RISCV_HWPROBE_MISALIGNED_EMULATED (1 << 0) diff --git a/arch/riscv/kernel/sys_hwprobe.c b/arch/riscv/kernel/sys_hwprobe.c index bcd3b816306c..d802ff707913 100644 --- a/arch/riscv/kernel/sys_hwprobe.c +++ b/arch/riscv/kernel/sys_hwprobe.c @@ -108,6 +108,8 @@ static void hwprobe_isa_ext0(struct riscv_hwprobe *pair, EXT_KEY(ZCB); EXT_KEY(ZCMOP); EXT_KEY(ZICBOZ); + EXT_KEY(ZICFILP); + EXT_KEY(ZICFISS); EXT_KEY(ZICOND); EXT_KEY(ZIHINTNTL); EXT_KEY(ZIHINTPAUSE);
From: Clément Léger cleger@rivosinc.com
Add necessary SBI definitions to use the FWFT extension.
Signed-off-by: Clément Léger cleger@rivosinc.com Reviewed-by: Zong Li zong.li@sifive.com --- arch/riscv/include/asm/sbi.h | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+)
diff --git a/arch/riscv/include/asm/sbi.h b/arch/riscv/include/asm/sbi.h index 3d250824178b..23bfb254e3f4 100644 --- a/arch/riscv/include/asm/sbi.h +++ b/arch/riscv/include/asm/sbi.h @@ -35,6 +35,7 @@ enum sbi_ext_id { SBI_EXT_DBCN = 0x4442434E, SBI_EXT_STA = 0x535441, SBI_EXT_NACL = 0x4E41434C, + SBI_EXT_FWFT = 0x46574654,
/* Experimentals extensions must lie within this range */ SBI_EXT_EXPERIMENTAL_START = 0x08000000, @@ -401,6 +402,31 @@ enum sbi_ext_nacl_feature {
#define SBI_NACL_SHMEM_SRET_X(__i) ((__riscv_xlen / 8) * (__i)) #define SBI_NACL_SHMEM_SRET_X_LAST 31 +/* SBI function IDs for FW feature extension */ +#define SBI_EXT_FWFT_SET 0x0 +#define SBI_EXT_FWFT_GET 0x1 + +enum sbi_fwft_feature_t { + SBI_FWFT_MISALIGNED_EXC_DELEG = 0x0, + SBI_FWFT_LANDING_PAD = 0x1, + SBI_FWFT_SHADOW_STACK = 0x2, + SBI_FWFT_DOUBLE_TRAP = 0x3, + SBI_FWFT_PTE_AD_HW_UPDATING = 0x4, + SBI_FWFT_LOCAL_RESERVED_START = 0x5, + SBI_FWFT_LOCAL_RESERVED_END = 0x3fffffff, + SBI_FWFT_LOCAL_PLATFORM_START = 0x40000000, + SBI_FWFT_LOCAL_PLATFORM_END = 0x7fffffff, + + SBI_FWFT_GLOBAL_RESERVED_START = 0x80000000, + SBI_FWFT_GLOBAL_RESERVED_END = 0xbfffffff, + SBI_FWFT_GLOBAL_PLATFORM_START = 0xc0000000, + SBI_FWFT_GLOBAL_PLATFORM_END = 0xffffffff, +}; + +#define SBI_FWFT_GLOBAL_FEATURE_BIT (1 << 31) +#define SBI_FWFT_PLATFORM_FEATURE_BIT (1 << 30) + +#define SBI_FWFT_SET_FLAG_LOCK (1 << 0)
/* SBI spec version fields */ #define SBI_SPEC_VERSION_DEFAULT 0x1
Kernel will have to perform shadow stack operations on user shadow stack. Like during signal delivery and sigreturn, shadow stack token must be created and validated respectively. Thus shadow stack access for kernel must be enabled.
In future when kernel shadow stacks are enabled for linux kernel, it must be enabled as early as possible for better coverage and prevent imbalance between regular stack and shadow stack. After `relocate_enable_mmu` has been done, this is as early as possible it can enabled.
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/kernel/asm-offsets.c | 4 ++++ arch/riscv/kernel/head.S | 12 ++++++++++++ 2 files changed, 16 insertions(+)
diff --git a/arch/riscv/kernel/asm-offsets.c b/arch/riscv/kernel/asm-offsets.c index 0c188aaf3925..21f99d5757b6 100644 --- a/arch/riscv/kernel/asm-offsets.c +++ b/arch/riscv/kernel/asm-offsets.c @@ -515,4 +515,8 @@ void asm_offsets(void) DEFINE(FREGS_A6, offsetof(struct __arch_ftrace_regs, a6)); DEFINE(FREGS_A7, offsetof(struct __arch_ftrace_regs, a7)); #endif + DEFINE(SBI_EXT_FWFT, SBI_EXT_FWFT); + DEFINE(SBI_EXT_FWFT_SET, SBI_EXT_FWFT_SET); + DEFINE(SBI_FWFT_SHADOW_STACK, SBI_FWFT_SHADOW_STACK); + DEFINE(SBI_FWFT_SET_FLAG_LOCK, SBI_FWFT_SET_FLAG_LOCK); } diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S index 356d5397b2a2..6244408ca917 100644 --- a/arch/riscv/kernel/head.S +++ b/arch/riscv/kernel/head.S @@ -164,6 +164,12 @@ secondary_start_sbi: call relocate_enable_mmu #endif call .Lsetup_trap_vector + li a7, SBI_EXT_FWFT + li a6, SBI_EXT_FWFT_SET + li a0, SBI_FWFT_SHADOW_STACK + li a1, 1 /* enable supervisor to access shadow stack access */ + li a2, SBI_FWFT_SET_FLAG_LOCK + ecall scs_load_current call smp_callin #endif /* CONFIG_SMP */ @@ -320,6 +326,12 @@ SYM_CODE_START(_start_kernel) la tp, init_task la sp, init_thread_union + THREAD_SIZE addi sp, sp, -PT_SIZE_ON_STACK + li a7, SBI_EXT_FWFT + li a6, SBI_EXT_FWFT_SET + li a0, SBI_FWFT_SHADOW_STACK + li a1, 1 /* enable supervisor to access shadow stack access */ + li a2, SBI_FWFT_SET_FLAG_LOCK + ecall scs_load_current
#ifdef CONFIG_KASAN
2025-03-14T14:39:41-07:00, Deepak Gupta debug@rivosinc.com:
Why can't kernel access the user shadow stack through an aliased WR mapping?
I think the ecall can fail even on machines that have Zicfiss, so it would be good to disable user shadow stack if that happens.
On Thu, Mar 20, 2025 at 3:10 PM Radim Krčmář rkrcmar@ventanamicro.com wrote:
It can, although that opens up a can of worms. If this alternating mapping is user mode then ensuring that another threat in userspace can't write to this address in this window of signal handling. A kernel alternate mapping can be created, but that can lead to all sorts of requirements of ensuring the page is pinned down. IIRC, It has been debated on x86 shadow stack merge time as well on how a flaky alias mapping approach can become and weaken the threat model it is supposed to protect against.
Simply using `ssamoswap` is simple and serves the purpose. Enabling shadow stack access for the kernel doesn't have any adverse effect on the kernel.
2025-03-20T15:42:44-07:00, Deepak Gupta debug@rivosinc.com:
Right, it must not be user mode.
True.
Simply using `ssamoswap` is simple and serves the purpose. Enabling shadow stack access for the kernel doesn't have any adverse effect on the kernel.
Makes sense. We just depend on an extra feature, because we should consider the case when M-mode doesn't allow S-mode shadow stack, but S-mode can enable U-mode shadow stack:
This commit adds a kernel command line option using which user cfi can be disabled.
Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/kernel/usercfi.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+)
diff --git a/arch/riscv/kernel/usercfi.c b/arch/riscv/kernel/usercfi.c index d31d89618763..813162ce4f15 100644 --- a/arch/riscv/kernel/usercfi.c +++ b/arch/riscv/kernel/usercfi.c @@ -17,6 +17,8 @@ #include <asm/csr.h> #include <asm/usercfi.h>
+bool disable_riscv_usercfi; + #define SHSTK_ENTRY_SIZE sizeof(void *)
bool is_shstk_enabled(struct task_struct *task) @@ -396,6 +398,9 @@ int arch_set_shadow_stack_status(struct task_struct *t, unsigned long status) unsigned long size = 0, addr = 0; bool enable_shstk = false;
+ if (disable_riscv_usercfi) + return 0; + if (!cpu_supports_shadow_stack()) return -EINVAL;
@@ -475,6 +480,9 @@ int arch_set_indir_br_lp_status(struct task_struct *t, unsigned long status) { bool enable_indir_lp = false;
+ if (disable_riscv_usercfi) + return 0; + if (!cpu_supports_indirect_br_lp_instr()) return -EINVAL;
@@ -507,3 +515,16 @@ int arch_lock_indir_br_lp_status(struct task_struct *task,
return 0; } + +static int __init setup_global_riscv_enable(char *str) +{ + if (strcmp(str, "true") == 0) + disable_riscv_usercfi = true; + + pr_info("Setting riscv usercfi to be %s\n", + (disable_riscv_usercfi ? "disabled" : "enabled")); + + return 1; +} + +__setup("disable_riscv_usercfi=", setup_global_riscv_enable);
2025-03-14T14:39:42-07:00, Deepak Gupta debug@rivosinc.com:
I'd prefer two command line options instead.
In any case, I think we still document params in kernel-parameters.txt.
On Thu, Mar 20, 2025 at 2:35 PM Radim Krčmář rkrcmar@ventanamicro.com wrote:
One for zicfilp and one for zicfiss ?
In any case, I think we still document params in kernel-parameters.txt.
Noted, will do that.
2025-03-20T15:31:09-07:00, Deepak Gupta debug@rivosinc.com:
Yes. I don't want to suggest the naming, because I would have used "nousercfi" without an argument for the current one.
From: Jim Shu jim.shu@sifive.com
user mode tasks compiled with zicfilp may call indirectly into vdso (like hwprobe indirect calls). Add landing pad compile support in vdso. vdso with landing pad in it will be nop for tasks which have not enabled landing pad. This patch allows to run user mode tasks with cfi eanbled and do no harm.
Future work can be done on this to do below - labeled landing pad on vdso functions (whenever labeling support shows up in gnu-toolchain) - emit shadow stack instructions only in vdso compiled objects as part of kernel compile.
Signed-off-by: Jim Shu jim.shu@sifive.com Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/Makefile | 5 +++- arch/riscv/include/asm/assembler.h | 44 +++++++++++++++++++++++++++++++++++ arch/riscv/kernel/vdso/Makefile | 12 ++++++++++ arch/riscv/kernel/vdso/flush_icache.S | 4 ++++ arch/riscv/kernel/vdso/getcpu.S | 4 ++++ arch/riscv/kernel/vdso/rt_sigreturn.S | 4 ++++ arch/riscv/kernel/vdso/sys_hwprobe.S | 4 ++++ 7 files changed, 76 insertions(+), 1 deletion(-)
diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile index 13fbc0f94238..eca94246cda6 100644 --- a/arch/riscv/Makefile +++ b/arch/riscv/Makefile @@ -88,9 +88,12 @@ riscv-march-$(CONFIG_TOOLCHAIN_HAS_ZACAS) := $(riscv-march-y)_zacas # Check if the toolchain supports Zabha riscv-march-$(CONFIG_TOOLCHAIN_HAS_ZABHA) := $(riscv-march-y)_zabha
+KBUILD_BASE_ISA = -march=$(shell echo $(riscv-march-y) | sed -E 's/(rv32ima|rv64ima)fd([^v_]*)v?/\1\2/') +export KBUILD_BASE_ISA + # Remove F,D,V from isa string for all. Keep extensions between "fd" and "v" by # matching non-v and non-multi-letter extensions out with the filter ([^v_]*) -KBUILD_CFLAGS += -march=$(shell echo $(riscv-march-y) | sed -E 's/(rv32ima|rv64ima)fd([^v_]*)v?/\1\2/') +KBUILD_CFLAGS += $(KBUILD_BASE_ISA)
KBUILD_AFLAGS += -march=$(riscv-march-y)
diff --git a/arch/riscv/include/asm/assembler.h b/arch/riscv/include/asm/assembler.h index 44b1457d3e95..a058ea5e9c58 100644 --- a/arch/riscv/include/asm/assembler.h +++ b/arch/riscv/include/asm/assembler.h @@ -80,3 +80,47 @@ .endm
#endif /* __ASM_ASSEMBLER_H */ + +#if defined(CONFIG_RISCV_USER_CFI) && (__riscv_xlen == 64) +.macro vdso_lpad +lpad 0 +.endm +#else +.macro vdso_lpad +.endm +#endif + +/* + * This macro emits a program property note section identifying + * architecture features which require special handling, mainly for + * use in assembly files included in the VDSO. + */ +#define NT_GNU_PROPERTY_TYPE_0 5 +#define GNU_PROPERTY_RISCV_FEATURE_1_AND 0xc0000000 + +#define GNU_PROPERTY_RISCV_FEATURE_1_ZICFILP (1U << 0) +#define GNU_PROPERTY_RISCV_FEATURE_1_ZICFISS (1U << 1) + +#if defined(CONFIG_RISCV_USER_CFI) && (__riscv_xlen == 64) +#define GNU_PROPERTY_RISCV_FEATURE_1_DEFAULT \ + (GNU_PROPERTY_RISCV_FEATURE_1_ZICFILP) +#endif + +#ifdef GNU_PROPERTY_RISCV_FEATURE_1_DEFAULT +.macro emit_riscv_feature_1_and, feat = GNU_PROPERTY_RISCV_FEATURE_1_DEFAULT + .pushsection .note.gnu.property, "a" + .p2align 3 + .word 4 + .word 16 + .word NT_GNU_PROPERTY_TYPE_0 + .asciz "GNU" + .word GNU_PROPERTY_RISCV_FEATURE_1_AND + .word 4 + .word \feat + .word 0 + .popsection +.endm +#else +.macro emit_riscv_feature_1_and, feat = 0 +.endm +#endif diff --git a/arch/riscv/kernel/vdso/Makefile b/arch/riscv/kernel/vdso/Makefile index 9a1b555e8733..daa10c2b0dd1 100644 --- a/arch/riscv/kernel/vdso/Makefile +++ b/arch/riscv/kernel/vdso/Makefile @@ -13,12 +13,18 @@ vdso-syms += flush_icache vdso-syms += hwprobe vdso-syms += sys_hwprobe
+ifdef CONFIG_RISCV_USER_CFI +LPAD_MARCH = _zicfilp +endif + # Files to link into the vdso obj-vdso = $(patsubst %, %.o, $(vdso-syms)) note.o
ccflags-y := -fno-stack-protector ccflags-y += -DDISABLE_BRANCH_PROFILING ccflags-y += -fno-builtin +ccflags-y += $(KBUILD_BASE_ISA)$(LPAD_MARCH) +asflags-y += $(KBUILD_BASE_ISA)$(LPAD_MARCH)
ifneq ($(c-gettimeofday-y),) CFLAGS_vgettimeofday.o += -fPIC -include $(c-gettimeofday-y) @@ -40,6 +46,12 @@ endif CFLAGS_REMOVE_vgettimeofday.o = $(CC_FLAGS_FTRACE) $(CC_FLAGS_SCS) CFLAGS_REMOVE_hwprobe.o = $(CC_FLAGS_FTRACE) $(CC_FLAGS_SCS)
+# Disable profiling and instrumentation for VDSO code +GCOV_PROFILE := n +KCOV_INSTRUMENT := n +KASAN_SANITIZE := n +UBSAN_SANITIZE := n + # Force dependency $(obj)/vdso.o: $(obj)/vdso.so
diff --git a/arch/riscv/kernel/vdso/flush_icache.S b/arch/riscv/kernel/vdso/flush_icache.S index 8f884227e8bc..e4c56970905e 100644 --- a/arch/riscv/kernel/vdso/flush_icache.S +++ b/arch/riscv/kernel/vdso/flush_icache.S @@ -5,11 +5,13 @@
#include <linux/linkage.h> #include <asm/unistd.h> +#include <asm/assembler.h>
.text /* int __vdso_flush_icache(void *start, void *end, unsigned long flags); */ SYM_FUNC_START(__vdso_flush_icache) .cfi_startproc + vdso_lpad #ifdef CONFIG_SMP li a7, __NR_riscv_flush_icache ecall @@ -20,3 +22,5 @@ SYM_FUNC_START(__vdso_flush_icache) ret .cfi_endproc SYM_FUNC_END(__vdso_flush_icache) + +emit_riscv_feature_1_and diff --git a/arch/riscv/kernel/vdso/getcpu.S b/arch/riscv/kernel/vdso/getcpu.S index 9c1bd531907f..5c1ecc4e1465 100644 --- a/arch/riscv/kernel/vdso/getcpu.S +++ b/arch/riscv/kernel/vdso/getcpu.S @@ -5,14 +5,18 @@
#include <linux/linkage.h> #include <asm/unistd.h> +#include <asm/assembler.h>
.text /* int __vdso_getcpu(unsigned *cpu, unsigned *node, void *unused); */ SYM_FUNC_START(__vdso_getcpu) .cfi_startproc + vdso_lpad /* For now, just do the syscall. */ li a7, __NR_getcpu ecall ret .cfi_endproc SYM_FUNC_END(__vdso_getcpu) + +emit_riscv_feature_1_and diff --git a/arch/riscv/kernel/vdso/rt_sigreturn.S b/arch/riscv/kernel/vdso/rt_sigreturn.S index 3dc022aa8931..e82987dc3739 100644 --- a/arch/riscv/kernel/vdso/rt_sigreturn.S +++ b/arch/riscv/kernel/vdso/rt_sigreturn.S @@ -5,12 +5,16 @@
#include <linux/linkage.h> #include <asm/unistd.h> +#include <asm/assembler.h>
.text SYM_FUNC_START(__vdso_rt_sigreturn) .cfi_startproc .cfi_signal_frame + vdso_lpad li a7, __NR_rt_sigreturn ecall .cfi_endproc SYM_FUNC_END(__vdso_rt_sigreturn) + +emit_riscv_feature_1_and diff --git a/arch/riscv/kernel/vdso/sys_hwprobe.S b/arch/riscv/kernel/vdso/sys_hwprobe.S index 77e57f830521..f1694451a60c 100644 --- a/arch/riscv/kernel/vdso/sys_hwprobe.S +++ b/arch/riscv/kernel/vdso/sys_hwprobe.S @@ -3,13 +3,17 @@
#include <linux/linkage.h> #include <asm/unistd.h> +#include <asm/assembler.h>
.text SYM_FUNC_START(riscv_hwprobe) .cfi_startproc + vdso_lpad li a7, __NR_riscv_hwprobe ecall ret
.cfi_endproc SYM_FUNC_END(riscv_hwprobe) + +emit_riscv_feature_1_and
On 14/03/2025 22:39, Deepak Gupta wrote:
So I looked into this and it seems it is not needed. All the profilers above only add their corresponding "clags" if the object "is-kernel-object" (https://elixir.bootlin.com/linux/v6.14-rc6/source/scripts/Makefile.lib#L147) which is not the case for hwprobe.c since it is not added to obj-y.
This patch creates a config for shadow stack support and landing pad instr support. Shadow stack support and landing instr support can be enabled by selecting `CONFIG_RISCV_USER_CFI`. Selecting `CONFIG_RISCV_USER_CFI` wires up path to enumerate CPU support and if cpu support exists, kernel will support cpu assisted user mode cfi.
If CONFIG_RISCV_USER_CFI is selected, select `ARCH_USES_HIGH_VMA_FLAGS`, `ARCH_HAS_USER_SHADOW_STACK` and DYNAMIC_SIGFRAME for riscv.
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- arch/riscv/Kconfig | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+)
diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig index 7612c52e9b1e..0a2e50f056e8 100644 --- a/arch/riscv/Kconfig +++ b/arch/riscv/Kconfig @@ -250,6 +250,26 @@ config ARCH_HAS_BROKEN_DWARF5 # https://github.com/llvm/llvm-project/commit/7ffabb61a5569444b5ac9322e22e5471... depends on LD_IS_LLD && LLD_VERSION < 180000
+config RISCV_USER_CFI + def_bool y + bool "riscv userspace control flow integrity" + depends on 64BIT && $(cc-option,-mabi=lp64 -march=rv64ima_zicfiss) + depends on RISCV_ALTERNATIVE + select ARCH_HAS_USER_SHADOW_STACK + select ARCH_USES_HIGH_VMA_FLAGS + select DYNAMIC_SIGFRAME + help + Provides CPU assisted control flow integrity to userspace tasks. + Control flow integrity is provided by implementing shadow stack for + backward edge and indirect branch tracking for forward edge in program. + Shadow stack protection is a hardware feature that detects function + return address corruption. This helps mitigate ROP attacks. + Indirect branch tracking enforces that all indirect branches must land + on a landing pad instruction else CPU will fault. This mitigates against + JOP / COP attacks. Applications must be enabled to use it, and old user- + space does not get protection "for free". + default y + config ARCH_MMAP_RND_BITS_MIN default 18 if 64BIT default 8
2025-03-14T14:39:44-07:00, Deepak Gupta debug@rivosinc.com:
A high level question to kick off my review:
Why are landing pads and shadow stacks merged together?
Apart from adding build flexibility, we could also split the patches into two isolated series, because the features are independent.
On Thu, Mar 20, 2025 at 2:25 PM Radim Krčmář rkrcmar@ventanamicro.com wrote:
Strictly from CPU extensions point of view they are independent features. Although from a usability point of view they complement each other. A user wanting to enable support for control flow integrity wouldn't be enabling only landing pad and leaving return flow open for an attacker and vice-versa. That's why I defined a single CONFIG for CFI.
From organizing patches in the patch series, shadow stack and landing pad patches do not cross into each other and are different from each other except dt-bindings, hwprobe, csr definitions. I can separate them out as well if that's desired.
Furthermore, I do not see an implementation only implementing zicfilp while not implementing zicfiss. There is a case of a nommu case where only zicfilp might be implemented and no zicfiss. However that's the case which is anyways riscv linux kernel is not actively being tested. IIRC, it was (nommu linux) considered to be phased out/not supported as well.
We could have two different configs but I don't see what would serve apart from the ability to build support for landing pad and shadow stack differently. As I said from a usability point of view both features are complimenting to each other rather than standing out alone and providing full protection.
A kernel is built with support for enabling both features or none. Sure user can use either of the prctl to enable either of the features in whatever combination they see fit.
2025-03-20T15:29:55-07:00, Deepak Gupta debug@rivosinc.com:
It would be my preference, but it's the maintainer's call.
I think that landing pads could be merged earlier if they were posted separately, but this series is already on v12, so I don't want to force anything.
Yeah, it will be rare, but if for some reason one feature cannot be used, then having just one is better than none. We can easily add compile options later and if we start with separate kernel parameters, then the users won't even notice a difference.
Adding documentation on landing pad aka indirect branch tracking on riscv and kernel interfaces exposed so that user tasks can enable it.
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- Documentation/arch/riscv/index.rst | 1 + Documentation/arch/riscv/zicfilp.rst | 115 +++++++++++++++++++++++++++++++++++ 2 files changed, 116 insertions(+)
diff --git a/Documentation/arch/riscv/index.rst b/Documentation/arch/riscv/index.rst index eecf347ce849..be7237b69682 100644 --- a/Documentation/arch/riscv/index.rst +++ b/Documentation/arch/riscv/index.rst @@ -14,6 +14,7 @@ RISC-V architecture uabi vector cmodx + zicfilp
features
diff --git a/Documentation/arch/riscv/zicfilp.rst b/Documentation/arch/riscv/zicfilp.rst new file mode 100644 index 000000000000..a188d78fcde6 --- /dev/null +++ b/Documentation/arch/riscv/zicfilp.rst @@ -0,0 +1,115 @@ +.. SPDX-License-Identifier: GPL-2.0 + +:Author: Deepak Gupta debug@rivosinc.com +:Date: 12 January 2024 + +==================================================== +Tracking indirect control transfers on RISC-V Linux +==================================================== + +This document briefly describes the interface provided to userspace by Linux +to enable indirect branch tracking for user mode applications on RISV-V + +1. Feature Overview +-------------------- + +Memory corruption issues usually result in to crashes, however when in hands of +an adversary and if used creatively can result into variety security issues. + +One of those security issues can be code re-use attacks on program where adversary +can use corrupt function pointers and chain them together to perform jump oriented +programming (JOP) or call oriented programming (COP) and thus compromising control +flow integrity (CFI) of the program. + +Function pointers live in read-write memory and thus are susceptible to corruption +and allows an adversary to reach any program counter (PC) in address space. On +RISC-V zicfilp extension enforces a restriction on such indirect control +transfers: + +- indirect control transfers must land on a landing pad instruction ``lpad``. + There are two exception to this rule: + + - rs1 = x1 or rs1 = x5, i.e. a return from a function and returns are + protected using shadow stack (see zicfiss.rst) + + - rs1 = x7. On RISC-V compiler usually does below to reach function + which is beyond the offset possible J-type instruction:: + + auipc x7, <imm> + jalr (x7) + + Such form of indirect control transfer are still immutable and don't rely + on memory and thus rs1=x7 is exempted from tracking and considered software + guarded jumps. + +``lpad`` instruction is pseudo of ``auipc rd, <imm_20bit>`` with ``rd=x0`` and +is a HINT nop. ``lpad`` instruction must be aligned on 4 byte boundary and +compares 20 bit immediate withx7. If ``imm_20bit`` == 0, CPU don't perform any +comparision with ``x7``. If ``imm_20bit`` != 0, then ``imm_20bit`` must match +``x7`` else CPU will raise ``software check exception`` (``cause=18``) with +``*tval = 2``. + +Compiler can generate a hash over function signatures and setup them (truncated +to 20bit) in x7 at callsites and function prologues can have ``lpad`` with same +function hash. This further reduces number of program counters a call site can +reach. + +2. ELF and psABI +----------------- + +Toolchain sets up :c:macro:`GNU_PROPERTY_RISCV_FEATURE_1_FCFI` for property +:c:macro:`GNU_PROPERTY_RISCV_FEATURE_1_AND` in notes section of the object file. + +3. Linux enabling +------------------ + +User space programs can have multiple shared objects loaded in its address space +and it's a difficult task to make sure all the dependencies have been compiled +with support of indirect branch. Thus it's left to dynamic loader to enable +indirect branch tracking for the program. + +4. prctl() enabling +-------------------- + +:c:macro:`PR_SET_INDIR_BR_LP_STATUS` / :c:macro:`PR_GET_INDIR_BR_LP_STATUS` / +:c:macro:`PR_LOCK_INDIR_BR_LP_STATUS` are three prctls added to manage indirect +branch tracking. prctls are arch agnostic and returns -EINVAL on other arches. + +* prctl(PR_SET_INDIR_BR_LP_STATUS, unsigned long arg) + +If arg1 is :c:macro:`PR_INDIR_BR_LP_ENABLE` and if CPU supports ``zicfilp`` +then kernel will enabled indirect branch tracking for the task. Dynamic loader +can issue this :c:macro:`prctl` once it has determined that all the objects +loaded in address space support indirect branch tracking. Additionally if there +is a `dlopen` to an object which wasn't compiled with ``zicfilp``, dynamic +loader can issue this prctl with arg1 set to 0 (i.e. +:c:macro:`PR_INDIR_BR_LP_ENABLE` being clear) + +* prctl(PR_GET_INDIR_BR_LP_STATUS, unsigned long arg) + +Returns current status of indirect branch tracking. If enabled it'll return +:c:macro:`PR_INDIR_BR_LP_ENABLE` + +* prctl(PR_LOCK_INDIR_BR_LP_STATUS, unsigned long arg) + +Locks current status of indirect branch tracking on the task. User space may +want to run with strict security posture and wouldn't want loading of objects +without ``zicfilp`` support in it and thus would want to disallow disabling of +indirect branch tracking. In that case user space can use this prctl to lock +current settings. + +5. violations related to indirect branch tracking +-------------------------------------------------- + +Pertaining to indirect branch tracking, CPU raises software check exception in +following conditions: + +- missing ``lpad`` after indirect call / jmp +- ``lpad`` not on 4 byte boundary +- ``imm_20bit`` embedded in ``lpad`` instruction doesn't match with ``x7`` + +In all 3 cases, ``*tval = 2`` is captured and software check exception is +raised (``cause=18``) + +Linux kernel will treat this as :c:macro:`SIGSEV`` with code = +:c:macro:`SEGV_CPERR` and follow normal course of signal delivery.
On 14/03/2025 22:39, Deepak Gupta wrote:
s/RISV-V/RISC-V and add a dot at the end of the line.
s/in to/into?
+an adversary and if used creatively can result into variety security issues.
"into a variety of security issues" ?
s/withx7/with x7
s/don't/doesn't
s/will enabled/will enable
Adding documentation on shadow stack for user mode on riscv and kernel interfaces exposed so that user tasks can enable it.
Reviewed-by: Zong Li zong.li@sifive.com Signed-off-by: Deepak Gupta debug@rivosinc.com --- Documentation/arch/riscv/index.rst | 1 + Documentation/arch/riscv/zicfiss.rst | 176 +++++++++++++++++++++++++++++++++++ 2 files changed, 177 insertions(+)
diff --git a/Documentation/arch/riscv/index.rst b/Documentation/arch/riscv/index.rst index be7237b69682..e240eb0ceb70 100644 --- a/Documentation/arch/riscv/index.rst +++ b/Documentation/arch/riscv/index.rst @@ -15,6 +15,7 @@ RISC-V architecture vector cmodx zicfilp + zicfiss
features
diff --git a/Documentation/arch/riscv/zicfiss.rst b/Documentation/arch/riscv/zicfiss.rst new file mode 100644 index 000000000000..5ba389f15b3f --- /dev/null +++ b/Documentation/arch/riscv/zicfiss.rst @@ -0,0 +1,176 @@ +.. SPDX-License-Identifier: GPL-2.0 + +:Author: Deepak Gupta debug@rivosinc.com +:Date: 12 January 2024 + +========================================================= +Shadow stack to protect function returns on RISC-V Linux +========================================================= + +This document briefly describes the interface provided to userspace by Linux +to enable shadow stack for user mode applications on RISV-V + +1. Feature Overview +-------------------- + +Memory corruption issues usually result in to crashes, however when in hands of +an adversary and if used creatively can result into variety security issues. + +One of those security issues can be code re-use attacks on program where +adversary can use corrupt return addresses present on stack and chain them +together to perform return oriented programming (ROP) and thus compromising +control flow integrity (CFI) of the program. + +Return addresses live on stack and thus in read-write memory and thus are +susceptible to corruption and allows an adversary to reach any program counter +(PC) in address space. On RISC-V ``zicfiss`` extension provides an alternate +stack termed as shadow stack on which return addresses can be safely placed in +prolog of the function and retrieved in epilog. ``zicfiss`` extension makes +following changes: + +- PTE encodings for shadow stack virtual memory + An earlier reserved encoding in first stage translation i.e. + PTE.R=0, PTE.W=1, PTE.X=0 becomes PTE encoding for shadow stack pages. + +- ``sspush x1/x5`` instruction pushes (stores) ``x1/x5`` to shadow stack. + +- ``sspopchk x1/x5`` instruction pops (loads) from shadow stack and compares + with ``x1/x5`` and if un-equal, CPU raises ``software check exception`` with + ``*tval = 3`` + +Compiler toolchain makes sure that function prologue have ``sspush x1/x5`` to +save return address on shadow stack in addition to regular stack. Similarly +function epilogs have ``ld x5, offset(x2)`` followed by ``sspopchk x5`` to +ensure that popped value from regular stack matches with popped value from +shadow stack. + +2. Shadow stack protections and linux memory manager +----------------------------------------------------- + +As mentioned earlier, shadow stack get new page table encodings and thus have +some special properties assigned to them and instructions that operate on them +as below: + +- Regular stores to shadow stack memory raises access store faults. This way + shadow stack memory is protected from stray inadvertant writes. + +- Regular loads to shadow stack memory are allowed. This allows stack trace + utilities or backtrace functions to read true callstack (not tampered). + +- Only shadow stack instructions can generate shadow stack load or shadow stack + store. + +- Shadow stack load / shadow stack store on read-only memory raises AMO/store + page fault. Thus both ``sspush x1/x5`` and ``sspopchk x1/x5`` will raise AMO/ + store page fault. This simplies COW handling in kernel During fork, kernel + can convert shadow stack pages into read-only memory (as it does for regular + read-write memory) and as soon as subsequent ``sspush`` or ``sspopchk`` in + userspace is encountered, then kernel can perform COW. + +- Shadow stack load / shadow stack store on read-write, read-write-execute + memory raises an access fault. This is a fatal condition because shadow stack + should never be operating on read-write, read-write-execute memory. + +3. ELF and psABI +----------------- + +Toolchain sets up :c:macro:`GNU_PROPERTY_RISCV_FEATURE_1_BCFI` for property +:c:macro:`GNU_PROPERTY_RISCV_FEATURE_1_AND` in notes section of the object file. + +4. Linux enabling +------------------ + +User space programs can have multiple shared objects loaded in its address space +and it's a difficult task to make sure all the dependencies have been compiled +with support of shadow stack. Thus it's left to dynamic loader to enable +shadow stack for the program. + +5. prctl() enabling +-------------------- + +:c:macro:`PR_SET_SHADOW_STACK_STATUS` / :c:macro:`PR_GET_SHADOW_STACK_STATUS` / +:c:macro:`PR_LOCK_SHADOW_STACK_STATUS` are three prctls added to manage shadow +stack enabling for tasks. prctls are arch agnostic and returns -EINVAL on other +arches. + +* prctl(PR_SET_SHADOW_STACK_STATUS, unsigned long arg) + +If arg1 :c:macro:`PR_SHADOW_STACK_ENABLE` and if CPU supports ``zicfiss`` then +kernel will enable shadow stack for the task. Dynamic loader can issue this +:c:macro:`prctl` once it has determined that all the objects loaded in address +space have support for shadow stack. Additionally if there is a +:c:macro:`dlopen` to an object which wasn't compiled with ``zicfiss``, dynamic +loader can issue this prctl with arg1 set to 0 (i.e. +:c:macro:`PR_SHADOW_STACK_ENABLE` being clear) + +* prctl(PR_GET_SHADOW_STACK_STATUS, unsigned long *arg) + +Returns current status of indirect branch tracking. If enabled it'll return +:c:macro:`PR_SHADOW_STACK_ENABLE`. + +* prctl(PR_LOCK_SHADOW_STACK_STATUS, unsigned long arg) + +Locks current status of shadow stack enabling on the task. User space may want +to run with strict security posture and wouldn't want loading of objects +without ``zicfiss`` support in it and thus would want to disallow disabling of +shadow stack on current task. In that case user space can use this prctl to +lock current settings. + +5. violations related to returns with shadow stack enabled +----------------------------------------------------------- + +Pertaining to shadow stack, CPU raises software check exception in following +condition: + +- On execution of ``sspopchk x1/x5``, ``x1/x5`` didn't match top of shadow + stack. If mismatch happens then cpu does ``*tval = 3`` and raise software + check exception. + +Linux kernel will treat this as :c:macro:`SIGSEV`` with code = +:c:macro:`SEGV_CPERR` and follow normal course of signal delivery. + +6. Shadow stack tokens +----------------------- +Regular stores on shadow stacks are not allowed and thus can't be tampered +with via arbitrary stray writes due to bugs. Method of pivoting / switching to +shadow stack is simply writing to csr ``CSR_SSP`` changes active shadow stack. +This can be problematic because usually value to be written to ``CSR_SSP`` will +be loaded somewhere in writeable memory and thus allows an adversary to +corruption bug in software to pivot to an any address in shadow stack range. +Shadow stack tokens can help mitigate this problem by making sure that: + +- When software is switching away from a shadow stack, shadow stack pointer + should be saved on shadow stack itself and call it ``shadow stack token`` + +- When software is switching to a shadow stack, it should read the + ``shadow stack token`` from shadow stack pointer and verify that + ``shadow stack token`` itself is pointer to shadow stack itself. + +- Once the token verification is done, software can perform the write to + ``CSR_SSP`` to switch shadow stack. + +Here software can be user mode task runtime itself which is managing various +contexts as part of single thread. Software can be kernel as well when kernel +has to deliver a signal to user task and must save shadow stack pointer. Kernel +can perform similar procedure by saving a token on user shadow stack itself. +This way whenever :c:macro:`sigreturn` happens, kernel can read the token and +verify the token and then switch to shadow stack. Using this mechanism, kernel +helps user task so that any corruption issue in user task is not exploited by +adversary by arbitrarily using :c:macro:`sigreturn`. Adversary will have to +make sure that there is a ``shadow stack token`` in addition to invoking +:c:macro:`sigreturn` + +7. Signal shadow stack +----------------------- +Following structure has been added to sigcontext for RISC-V:: + + struct __sc_riscv_cfi_state { + unsigned long ss_ptr; + }; + +As part of signal delivery, shadow stack token is saved on current shadow stack +itself and updated pointer is saved away in :c:macro:`ss_ptr` field in +:c:macro:`__sc_riscv_cfi_state` under :c:macro:`sigcontext`. Existing shadow +stack allocation is used for signal delivery. During :c:macro:`sigreturn`, +kernel will obtain :c:macro:`ss_ptr` from :c:macro:`sigcontext` and verify the +saved token on shadow stack itself and switch shadow stack.
On 14/03/2025 22:39, Deepak Gupta wrote:
s/RISV-V/RISC-V and the final dot
s/in to/into
+an adversary and if used creatively can result into variety security issues.
"into a variety of"
s/and allows/which allows
s/shadow stack/shadow stacks
s/inadvertant/inadvertent
s/During/during
I don't understand the end of this sentence.
Remove "an"
Adds kselftest for RISC-V control flow integrity implementation for user mode. There is not a lot going on in kernel for enabling landing pad for user mode. cfi selftest are intended to be compiled with zicfilp and zicfiss enabled compiler. Thus kselftest simply checks if landing pad and shadow stack for the binary and process are enabled or not. selftest then register a signal handler for SIGSEGV. Any control flow violation are reported as SIGSEGV with si_code = SEGV_CPERR. Test will fail on receiving any SEGV_CPERR. Shadow stack part has more changes in kernel and thus there are separate tests for that
- Exercise `map_shadow_stack` syscall - `fork` test to make sure COW works for shadow stack pages - gup tests Kernel uses FOLL_FORCE when access happens to memory via /proc/<pid>/mem. Not breaking that for shadow stack. - signal test. Make sure signal delivery results in token creation on shadow stack and consumes (and verifies) token on sigreturn - shadow stack protection test. attempts to write using regular store instruction on shadow stack memory must result in access faults
Test outut ==========
""" TAP version 13 1..5 This is to ensure shadow stack is indeed enabled and working This is to ensure shadow stack is indeed enabled and working ok 1 shstk fork test ok 2 map shadow stack syscall ok 3 shadow stack gup tests ok 4 shadow stack signal tests ok 5 memory protections of shadow stack memory """
Signed-off-by: Deepak Gupta debug@rivosinc.com --- tools/testing/selftests/riscv/Makefile | 2 +- tools/testing/selftests/riscv/cfi/.gitignore | 3 + tools/testing/selftests/riscv/cfi/Makefile | 10 + tools/testing/selftests/riscv/cfi/cfi_rv_test.h | 82 +++++ tools/testing/selftests/riscv/cfi/riscv_cfi_test.c | 78 +++++ tools/testing/selftests/riscv/cfi/shadowstack.c | 385 +++++++++++++++++++++ tools/testing/selftests/riscv/cfi/shadowstack.h | 27 ++ 7 files changed, 586 insertions(+), 1 deletion(-)
diff --git a/tools/testing/selftests/riscv/Makefile b/tools/testing/selftests/riscv/Makefile index 099b8c1f46f8..5671b4405a12 100644 --- a/tools/testing/selftests/riscv/Makefile +++ b/tools/testing/selftests/riscv/Makefile @@ -5,7 +5,7 @@ ARCH ?= $(shell uname -m 2>/dev/null || echo not)
ifneq (,$(filter $(ARCH),riscv)) -RISCV_SUBTARGETS ?= abi hwprobe mm sigreturn vector +RISCV_SUBTARGETS ?= abi hwprobe mm sigreturn vector cfi else RISCV_SUBTARGETS := endif diff --git a/tools/testing/selftests/riscv/cfi/.gitignore b/tools/testing/selftests/riscv/cfi/.gitignore new file mode 100644 index 000000000000..82545863bac6 --- /dev/null +++ b/tools/testing/selftests/riscv/cfi/.gitignore @@ -0,0 +1,3 @@ +cfitests +riscv_cfi_test +shadowstack diff --git a/tools/testing/selftests/riscv/cfi/Makefile b/tools/testing/selftests/riscv/cfi/Makefile new file mode 100644 index 000000000000..b65f7ff38a32 --- /dev/null +++ b/tools/testing/selftests/riscv/cfi/Makefile @@ -0,0 +1,10 @@ +CFLAGS += -I$(top_srcdir)/tools/include + +CFLAGS += -march=rv64gc_zicfilp_zicfiss + +TEST_GEN_PROGS := cfitests + +include ../../lib.mk + +$(OUTPUT)/cfitests: riscv_cfi_test.c shadowstack.c + $(CC) -o$@ $(CFLAGS) $(LDFLAGS) $^ diff --git a/tools/testing/selftests/riscv/cfi/cfi_rv_test.h b/tools/testing/selftests/riscv/cfi/cfi_rv_test.h new file mode 100644 index 000000000000..1c8043f2b778 --- /dev/null +++ b/tools/testing/selftests/riscv/cfi/cfi_rv_test.h @@ -0,0 +1,82 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ + +#ifndef SELFTEST_RISCV_CFI_H +#define SELFTEST_RISCV_CFI_H +#include <stddef.h> +#include <sys/types.h> +#include "shadowstack.h" + +#define CHILD_EXIT_CODE_SSWRITE 10 +#define CHILD_EXIT_CODE_SIG_TEST 11 + +#define my_syscall5(num, arg1, arg2, arg3, arg4, arg5) \ +({ \ + register long _num __asm__ ("a7") = (num); \ + register long _arg1 __asm__ ("a0") = (long)(arg1); \ + register long _arg2 __asm__ ("a1") = (long)(arg2); \ + register long _arg3 __asm__ ("a2") = (long)(arg3); \ + register long _arg4 __asm__ ("a3") = (long)(arg4); \ + register long _arg5 __asm__ ("a4") = (long)(arg5); \ + \ + __asm__ volatile( \ + "ecall\n" \ + : "+r" \ + (_arg1) \ + : "r"(_arg2), "r"(_arg3), "r"(_arg4), "r"(_arg5), \ + "r"(_num) \ + : "memory", "cc" \ + ); \ + _arg1; \ +}) + +#define my_syscall3(num, arg1, arg2, arg3) \ +({ \ + register long _num __asm__ ("a7") = (num); \ + register long _arg1 __asm__ ("a0") = (long)(arg1); \ + register long _arg2 __asm__ ("a1") = (long)(arg2); \ + register long _arg3 __asm__ ("a2") = (long)(arg3); \ + \ + __asm__ volatile( \ + "ecall\n" \ + : "+r" (_arg1) \ + : "r"(_arg2), "r"(_arg3), \ + "r"(_num) \ + : "memory", "cc" \ + ); \ + _arg1; \ +}) + +#ifndef __NR_prctl +#define __NR_prctl 167 +#endif + +#ifndef __NR_map_shadow_stack +#define __NR_map_shadow_stack 453 +#endif + +#define CSR_SSP 0x011 + +#ifdef __ASSEMBLY__ +#define __ASM_STR(x) x +#else +#define __ASM_STR(x) #x +#endif + +#define csr_read(csr) \ +({ \ + register unsigned long __v; \ + __asm__ __volatile__ ("csrr %0, " __ASM_STR(csr) \ + : "=r" (__v) : \ + : "memory"); \ + __v; \ +}) + +#define csr_write(csr, val) \ +({ \ + unsigned long __v = (unsigned long)(val); \ + __asm__ __volatile__ ("csrw " __ASM_STR(csr) ", %0" \ + : : "rK" (__v) \ + : "memory"); \ +}) + +#endif diff --git a/tools/testing/selftests/riscv/cfi/riscv_cfi_test.c b/tools/testing/selftests/riscv/cfi/riscv_cfi_test.c new file mode 100644 index 000000000000..cf33aa25cc73 --- /dev/null +++ b/tools/testing/selftests/riscv/cfi/riscv_cfi_test.c @@ -0,0 +1,78 @@ +// SPDX-License-Identifier: GPL-2.0-only + +#include "../../kselftest.h" +#include <sys/signal.h> +#include <asm/ucontext.h> +#include <linux/prctl.h> +#include "cfi_rv_test.h" + +/* do not optimize cfi related test functions */ +#pragma GCC push_options +#pragma GCC optimize("O0") + +void sigsegv_handler(int signum, siginfo_t *si, void *uc) +{ + struct ucontext *ctx = (struct ucontext *)uc; + + if (si->si_code == SEGV_CPERR) { + ksft_print_msg("Control flow violation happened somewhere\n"); + ksft_print_msg("PC where violation happened %lx\n", ctx->uc_mcontext.gregs[0]); + exit(-1); + } + + /* all other cases are expected to be of shadow stack write case */ + exit(CHILD_EXIT_CODE_SSWRITE); +} + +bool register_signal_handler(void) +{ + struct sigaction sa = {}; + + sa.sa_sigaction = sigsegv_handler; + sa.sa_flags = SA_SIGINFO; + if (sigaction(SIGSEGV, &sa, NULL)) { + ksft_print_msg("Registering signal handler for landing pad violation failed\n"); + return false; + } + + return true; +} + +int main(int argc, char *argv[]) +{ + int ret = 0; + unsigned long lpad_status = 0, ss_status = 0; + + ksft_print_header(); + + ksft_print_msg("Starting risc-v tests\n"); + + /* + * Landing pad test. Not a lot of kernel changes to support landing + * pad for user mode except lighting up a bit in senvcfg via a prctl + * Enable landing pad through out the execution of test binary + */ + ret = my_syscall5(__NR_prctl, PR_GET_INDIR_BR_LP_STATUS, &lpad_status, 0, 0, 0); + if (ret) + ksft_exit_fail_msg("Get landing pad status failed with %d\n", ret); + + if (!(lpad_status & PR_INDIR_BR_LP_ENABLE)) + ksft_exit_fail_msg("Landing pad is not enabled, should be enabled via glibc\n"); + + ret = my_syscall5(__NR_prctl, PR_GET_SHADOW_STACK_STATUS, &ss_status, 0, 0, 0); + if (ret) + ksft_exit_fail_msg("Get shadow stack failed with %d\n", ret); + + if (!(ss_status & PR_SHADOW_STACK_ENABLE)) + ksft_exit_fail_msg("Shadow stack is not enabled, should be enabled via glibc\n"); + + if (!register_signal_handler()) + ksft_exit_fail_msg("Registering signal handler for SIGSEGV failed\n"); + + ksft_print_msg("Landing pad and shadow stack are enabled for binary\n"); + execute_shadow_stack_tests(); + + return 0; +} + +#pragma GCC pop_options diff --git a/tools/testing/selftests/riscv/cfi/shadowstack.c b/tools/testing/selftests/riscv/cfi/shadowstack.c new file mode 100644 index 000000000000..53387dbd9cf5 --- /dev/null +++ b/tools/testing/selftests/riscv/cfi/shadowstack.c @@ -0,0 +1,385 @@ +// SPDX-License-Identifier: GPL-2.0-only + +#include "../../kselftest.h" +#include <sys/wait.h> +#include <signal.h> +#include <fcntl.h> +#include <asm-generic/unistd.h> +#include <sys/mman.h> +#include "shadowstack.h" +#include "cfi_rv_test.h" + +static struct shadow_stack_tests shstk_tests[] = { + { "shstk fork test\n", shadow_stack_fork_test }, + { "map shadow stack syscall\n", shadow_stack_map_test }, + { "shadow stack gup tests\n", shadow_stack_gup_tests }, + { "shadow stack signal tests\n", shadow_stack_signal_test}, + { "memory protections of shadow stack memory\n", shadow_stack_protection_test } +}; + +#define RISCV_SHADOW_STACK_TESTS ARRAY_SIZE(shstk_tests) + +/* do not optimize shadow stack related test functions */ +#pragma GCC push_options +#pragma GCC optimize("O0") + +void zar(void) +{ + unsigned long ssp = 0; + + ssp = csr_read(CSR_SSP); + ksft_print_msg("Spewing out shadow stack ptr: %lx\n" + " This is to ensure shadow stack is indeed enabled and working\n", + ssp); +} + +void bar(void) +{ + zar(); +} + +void foo(void) +{ + bar(); +} + +void zar_child(void) +{ + unsigned long ssp = 0; + + ssp = csr_read(CSR_SSP); + ksft_print_msg("Spewing out shadow stack ptr: %lx\n" + " This is to ensure shadow stack is indeed enabled and working\n", + ssp); +} + +void bar_child(void) +{ + zar_child(); +} + +void foo_child(void) +{ + bar_child(); +} + +typedef void (call_func_ptr)(void); +/* + * call couple of functions to test push pop. + */ +int shadow_stack_call_tests(call_func_ptr fn_ptr, bool parent) +{ + ksft_print_msg("dummy calls for sspush and sspopchk in context of %s\n", + parent ? "parent" : "child"); + + (fn_ptr)(); + + return 0; +} + +/* forks a thread, and ensure shadow stacks fork out */ +bool shadow_stack_fork_test(unsigned long test_num, void *ctx) +{ + int pid = 0, child_status = 0, parent_pid = 0, ret = 0; + unsigned long ss_status = 0; + + ksft_print_msg("Exercising shadow stack fork test\n"); + + ret = my_syscall5(__NR_prctl, PR_GET_SHADOW_STACK_STATUS, &ss_status, 0, 0, 0); + if (ret) { + ksft_exit_skip("Shadow stack get status prctl failed with errorcode %d\n", ret); + return false; + } + + if (!(ss_status & PR_SHADOW_STACK_ENABLE)) + ksft_exit_skip("Shadow stack is not enabled, should be enabled via glibc\n"); + + parent_pid = getpid(); + pid = fork(); + + if (pid) { + ksft_print_msg("Parent pid %d and child pid %d\n", parent_pid, pid); + shadow_stack_call_tests(&foo, true); + } else { + shadow_stack_call_tests(&foo_child, false); + } + + if (pid) { + ksft_print_msg("Waiting on child to finish\n"); + wait(&child_status); + } else { + /* exit child gracefully */ + exit(0); + } + + if (pid && WIFSIGNALED(child_status)) { + ksft_print_msg("Child faulted, fork test failed\n"); + return false; + } + + return true; +} + +/* exercise `map_shadow_stack`, pivot to it and call some functions to ensure it works */ +#define SHADOW_STACK_ALLOC_SIZE 4096 +bool shadow_stack_map_test(unsigned long test_num, void *ctx) +{ + unsigned long shdw_addr; + int ret = 0; + + ksft_print_msg("Exercising shadow stack map test\n"); + + shdw_addr = my_syscall3(__NR_map_shadow_stack, NULL, SHADOW_STACK_ALLOC_SIZE, 0); + + if (((long)shdw_addr) <= 0) { + ksft_print_msg("map_shadow_stack failed with error code %d\n", + (int)shdw_addr); + return false; + } + + ret = munmap((void *)shdw_addr, SHADOW_STACK_ALLOC_SIZE); + + if (ret) { + ksft_print_msg("munmap failed with error code %d\n", ret); + return false; + } + + return true; +} + +/* + * shadow stack protection tests. map a shadow stack and + * validate all memory protections work on it + */ +bool shadow_stack_protection_test(unsigned long test_num, void *ctx) +{ + unsigned long shdw_addr; + unsigned long *write_addr = NULL; + int ret = 0, pid = 0, child_status = 0; + + ksft_print_msg("Exercising shadow stack protection test (WPT)\n"); + + shdw_addr = my_syscall3(__NR_map_shadow_stack, NULL, SHADOW_STACK_ALLOC_SIZE, 0); + + if (((long)shdw_addr) <= 0) { + ksft_print_msg("map_shadow_stack failed with error code %d\n", + (int)shdw_addr); + return false; + } + + write_addr = (unsigned long *)shdw_addr; + pid = fork(); + + /* no child was created, return false */ + if (pid == -1) + return false; + + /* + * try to perform a store from child on shadow stack memory + * it should result in SIGSEGV + */ + if (!pid) { + /* below write must lead to SIGSEGV */ + *write_addr = 0xdeadbeef; + } else { + wait(&child_status); + } + + /* test fail, if 0xdeadbeef present on shadow stack address */ + if (*write_addr == 0xdeadbeef) { + ksft_print_msg("Shadow stack WPT failed\n"); + return false; + } + + /* if child reached here, then fail */ + if (!pid) { + ksft_print_msg("Shadow stack WPT failed: child reached unreachable state\n"); + return false; + } + + /* if child exited via signal handler but not for write on ss */ + if (WIFEXITED(child_status) && + WEXITSTATUS(child_status) != CHILD_EXIT_CODE_SSWRITE) { + ksft_print_msg("Shadow stack WPT failed: child wasn't signaled for write\n"); + return false; + } + + ret = munmap(write_addr, SHADOW_STACK_ALLOC_SIZE); + if (ret) { + ksft_print_msg("Shadow stack WPT failed: munmap failed, error code %d\n", + ret); + return false; + } + + return true; +} + +#define SS_MAGIC_WRITE_VAL 0xbeefdead + +int gup_tests(int mem_fd, unsigned long *shdw_addr) +{ + unsigned long val = 0; + + lseek(mem_fd, (unsigned long)shdw_addr, SEEK_SET); + if (read(mem_fd, &val, sizeof(val)) < 0) { + ksft_print_msg("Reading shadow stack mem via gup failed\n"); + return 1; + } + + val = SS_MAGIC_WRITE_VAL; + lseek(mem_fd, (unsigned long)shdw_addr, SEEK_SET); + if (write(mem_fd, &val, sizeof(val)) < 0) { + ksft_print_msg("Writing shadow stack mem via gup failed\n"); + return 1; + } + + if (*shdw_addr != SS_MAGIC_WRITE_VAL) { + ksft_print_msg("GUP write to shadow stack memory failed\n"); + return 1; + } + + return 0; +} + +bool shadow_stack_gup_tests(unsigned long test_num, void *ctx) +{ + unsigned long shdw_addr = 0; + unsigned long *write_addr = NULL; + int fd = 0; + bool ret = false; + + ksft_print_msg("Exercising shadow stack gup tests\n"); + shdw_addr = my_syscall3(__NR_map_shadow_stack, NULL, SHADOW_STACK_ALLOC_SIZE, 0); + + if (((long)shdw_addr) <= 0) { + ksft_print_msg("map_shadow_stack failed with error code %d\n", (int)shdw_addr); + return false; + } + + write_addr = (unsigned long *)shdw_addr; + + fd = open("/proc/self/mem", O_RDWR); + if (fd == -1) + return false; + + if (gup_tests(fd, write_addr)) { + ksft_print_msg("gup tests failed\n"); + goto out; + } + + ret = true; +out: + if (shdw_addr && munmap(write_addr, SHADOW_STACK_ALLOC_SIZE)) { + ksft_print_msg("munmap failed with error code %d\n", ret); + ret = false; + } + + return ret; +} + +volatile bool break_loop; + +void sigusr1_handler(int signo) +{ + break_loop = true; +} + +bool sigusr1_signal_test(void) +{ + struct sigaction sa = {}; + + sa.sa_handler = sigusr1_handler; + sa.sa_flags = 0; + sigemptyset(&sa.sa_mask); + if (sigaction(SIGUSR1, &sa, NULL)) { + ksft_print_msg("Registering signal handler for SIGUSR1 failed\n"); + return false; + } + + return true; +} + +/* + * shadow stack signal test. shadow stack must be enabled. + * register a signal, fork another thread which is waiting + * on signal. Send a signal from parent to child, verify + * that signal was received by child. If not test fails + */ +bool shadow_stack_signal_test(unsigned long test_num, void *ctx) +{ + int pid = 0, child_status = 0, ret = 0; + unsigned long ss_status = 0; + + ksft_print_msg("Exercising shadow stack signal test\n"); + + ret = my_syscall5(__NR_prctl, PR_GET_SHADOW_STACK_STATUS, &ss_status, 0, 0, 0); + if (ret) { + ksft_print_msg("Shadow stack get status prctl failed with errorcode %d\n", ret); + return false; + } + + if (!(ss_status & PR_SHADOW_STACK_ENABLE)) + ksft_print_msg("Shadow stack is not enabled, should be enabled via glibc\n"); + + /* this should be caught by signal handler and do an exit */ + if (!sigusr1_signal_test()) { + ksft_print_msg("Registering sigusr1 handler failed\n"); + exit(-1); + } + + pid = fork(); + + if (pid == -1) { + ksft_print_msg("Signal test: fork failed\n"); + goto out; + } + + if (pid == 0) { + while (!break_loop) + sleep(1); + + exit(11); + /* child shouldn't go beyond here */ + } + + /* send SIGUSR1 to child */ + kill(pid, SIGUSR1); + wait(&child_status); + +out: + + return (WIFEXITED(child_status) && + WEXITSTATUS(child_status) == 11); +} + +int execute_shadow_stack_tests(void) +{ + int ret = 0; + unsigned long test_count = 0; + unsigned long shstk_status = 0; + bool test_pass = false; + + ksft_print_msg("Executing RISC-V shadow stack self tests\n"); + ksft_set_plan(RISCV_SHADOW_STACK_TESTS); + + ret = my_syscall5(__NR_prctl, PR_GET_SHADOW_STACK_STATUS, &shstk_status, 0, 0, 0); + + if (ret != 0) + ksft_exit_fail_msg("Get shadow stack status failed with %d\n", ret); + + /* + * If we are here that means get shadow stack status succeeded and + * thus shadow stack support is baked in the kernel. + */ + while (test_count < RISCV_SHADOW_STACK_TESTS) { + test_pass = (*shstk_tests[test_count].t_func)(test_count, NULL); + ksft_test_result(test_pass, shstk_tests[test_count].name); + test_count++; + } + + ksft_finished(); + + return 0; +} + +#pragma GCC pop_options diff --git a/tools/testing/selftests/riscv/cfi/shadowstack.h b/tools/testing/selftests/riscv/cfi/shadowstack.h new file mode 100644 index 000000000000..0be510167de3 --- /dev/null +++ b/tools/testing/selftests/riscv/cfi/shadowstack.h @@ -0,0 +1,27 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ + +#ifndef SELFTEST_SHADOWSTACK_TEST_H +#define SELFTEST_SHADOWSTACK_TEST_H +#include <stddef.h> +#include <linux/prctl.h> + +/* + * a cfi test returns true for success or false for fail + * takes a number for test number to index into array and void pointer. + */ +typedef bool (*shstk_test_func)(unsigned long test_num, void *); + +struct shadow_stack_tests { + char *name; + shstk_test_func t_func; +}; + +bool shadow_stack_fork_test(unsigned long test_num, void *ctx); +bool shadow_stack_map_test(unsigned long test_num, void *ctx); +bool shadow_stack_protection_test(unsigned long test_num, void *ctx); +bool shadow_stack_gup_tests(unsigned long test_num, void *ctx); +bool shadow_stack_signal_test(unsigned long test_num, void *ctx); + +int execute_shadow_stack_tests(void); + +#endif
linux-kselftest-mirror@lists.linaro.org
-
Alexandre Ghiti -
Arnd Bergmann -
Deepak Gupta -
Radim Krčmář -
Zong Li