Initialize Vector registers in the first-use trap
cover Initialize Vector registers in the first-use trap
Before applying this series, We only initialize the space for saving Vector registers. This is not enough as Vector registers themselves also neeeded to be initialized before dropping back into userspace. Or, we may risk leaking a process's data left in Vector register to another process. This can be verified by only applying the second patch and running the test.
Andy Chiu (2):
riscv: vector: clear V-reg in the first-use trap
selftests: Test RISC-V Vector's first-use handler
arch/riscv/kernel/vector.c | 1 + .../testing/selftests/riscv/vector/.gitignore | 1 + tools/testing/selftests/riscv/vector/Makefile | 6 +- .../selftests/riscv/vector/v_initval_nolibc.c | 68 +++++++++++++++++++ 4 files changed, 75 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/riscv/vector/v_initval_nolibc.c
Signature · 7 lines
2.17.1 _______________________________________________ linux-riscv mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-riscv
PATCH 1/2 riscv: vector: clear V-reg in the first-use trap
If there is no context switch happens after we enable V for a process, then we return to user space with whatever left on the CPU's V registers accessible to the process. The leaked data could belong to another process's V-context saved from last context switch, impacting process's confidentiality on the system.
To prevent this from happening, we clear V registers by restoring zero'd V context after turining on V.
--- arch/riscv/kernel/vector.c | 1 + 1 file changed, 1 insertion(+)
arch/riscv/kernel/vector.c +1 -0
@@ -167,6 +167,7 @@ bool riscv_v_first_use_handler(struct pt_regs *regs)return true; }riscv_v_vstate_on (regs); +riscv_v_vstate_restore (current, regs);return true; }
Signature · 7 lines
2.17.1 _______________________________________________ linux-riscv mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-riscv
Re: [v1, 1/2] riscv: vector: clear V-reg in the first-use trap
Quoted 11 lines · Andy Chiu
If there is no context switch happens after we enable V for a process, then we return to user space with whatever left on the CPU's V registers accessible to the process. The leaked data could belong to another process's V-context saved from last context switch, impacting process's confidentiality on the system.
To prevent this from happening, we clear V registers by restoring zero'd V context after turining on V.
Fixes cd054837243b ("riscv: Allocate user's vector context in the first-use trap")Signed-off-by Andy Chiu <[email protected]>
_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
PATCH 2/2 selftests: Test RISC-V Vector's first-use handler
In reply to Andy ChiuThis add a test to check if the kernel zero-initializes all V registers after the first-use trap handler returns.
If V registers are not zero-initialized, then the test should fail one out of several runs:
``` root@sifive-fpga:~# ./v_initval_nolibc # vl = 256 not ok 1 detect stale values on v-regesters 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4c 41 4e 47 3d 43 0 50 41 54 48 3d 2f 75 73 72 2f 6c 6f 63 61 6c 2f 73 62 69 6e 3a 2f 75 73 72 2f 6c 6f 63 61 6c 2f 62 69 6e 3a 2f 75 73 72 ff ff 81 0 0 0 0 0 0 0 0 0 0 0 0 0 ```
Otherwise, the test passes without errors each run.
--- .../testing/selftests/riscv/vector/.gitignore | 1 + tools/testing/selftests/riscv/vector/Makefile | 6 +- .../selftests/riscv/vector/v_initval_nolibc.c | 68 +++++++++++++++++++ 3 files changed, 74 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/riscv/vector/v_initval_nolibc.c
tools/testing/selftests/riscv/vector/.gitignore +1 -0
@@ -1,2 +1,3 @@ vstate_exec_nolibc vstate_prctl +v_initval_nolibc
tools/testing/selftests/riscv/vector/Makefile +5 -1
@@ -2,7 +2,7 @@ # Copyright (C) 2021 ARM Limited # Originally tools/testing/arm64/abi/Makefile -TEST_GEN_PROGS := vstate_prctl +TEST_GEN_PROGS := vstate_prctl v_initval_nolibc TEST_GEN_PROGS_EXTENDED := vstate_exec_nolibc include ../../lib.mk
@@ -13,3 +13,7 @@ $(OUTPUT)/vstate_prctl: vstate_prctl.c ../hwprobe/sys_hwprobe.S $(OUTPUT)/vstate_exec_nolibc: vstate_exec_nolibc.c $(CC) -nostdlib -static -include ../../../../include/nolibc/nolibc.h \ -Wall $(CFLAGS) $(LDFLAGS) $^ -o $@ -lgcc + +$(OUTPUT)/v_initval_nolibc: v_initval_nolibc.c + $(CC) -nostdlib -static -include ../../../../include/nolibc/nolibc.h \ + -Wall $(CFLAGS) $(LDFLAGS) $^ -o $@ -lgcc
tools/testing/selftests/riscv/vector/v_initval_nolibc.c new file +68 -0
68 lines not on this page · raw message
Signature · 7 lines
2.17.1 _______________________________________________ linux-riscv mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-riscv
Re: [v1, 2/2] selftests: Test RISC-V Vector's first-use handler
Quoted 49 lines · Andy Chiu
This add a test to check if the kernel zero-initializes all V registers after the first-use trap handler returns.
If V registers are not zero-initialized, then the test should fail one out of several runs:
``` root@sifive-fpga:~# ./v_initval_nolibc # vl = 256 not ok 1 detect stale values on v-regesters 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4c 41 4e 47 3d 43 0 50 41 54 48 3d 2f 75 73 72 2f 6c 6f 63 61 6c 2f 73 62 69 6e 3a 2f 75 73 72 2f 6c 6f 63 61 6c 2f 62 69 6e 3a 2f 75 73 72 ff ff 81 0 0 0 0 0 0 0 0 0 0 0 0 0 ```Otherwise, the test passes without errors each run.
Signed-off-by Andy Chiu <[email protected]>--- .../testing/selftests/riscv/vector/.gitignore | 1 + tools/testing/selftests/riscv/vector/Makefile | 6 +- .../selftests/riscv/vector/v_initval_nolibc.c | 68 +++++++++++++++++++ 3 files changed, 74 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/riscv/vector/v_initval_nolibc.ctools/testing/selftests/riscv/vector/.gitignore +1
-0@@ -1,2 +1,3 @@ vstate_exec_nolibc vstate_prctl +v_initval_nolibctools/testing/selftests/riscv/vector/Makefile +1
-1@@ -2,7 +2,7 @@ # Copyright (C) 2021 ARM Limited # Originally tools/testing/arm64/abi/Makefile -TEST_GEN_PROGS := vstate_prctl +TEST_GEN_PROGS := vstate_prctl v_initval_nolibc TEST_GEN_PROGS_EXTENDED := vstate_exec_nolibc include ../../lib.mk
tools/testing/selftests/riscv/vector/Makefile $(OUTPUT)/vstate_prctl: vstate_prctl.c ../hwprobe/sys_hwprobe.S · line 19
Quoted 7 lines
tools/testing/selftests/riscv/vector/Makefile +4
-0@@ -13,3 +13,7 @@ $(OUTPUT)/vstate_prctl: vstate_prctl.c ../hwprobe/sys_hwprobe.S $(OUTPUT)/vstate_exec_nolibc: vstate_exec_nolibc.c $(CC) -nostdlib -static -include ../../../../include/nolibc/nolibc.h \ -Wall $(CFLAGS) $(LDFLAGS) $^ -o $@ -lgcc + +$(OUTPUT)/v_initval_nolibc: v_initval_nolibc.c + $(CC) -nostdlib -static -include ../../../../include/nolibc/nolibc.h \ + -Wall $(CFLAGS) $(LDFLAGS) $^ -o $@ -lgcc
Hmm, does this build with clang? (No, bigge on my end, and can be fixed later.)
Quoted 48 lines
tools/testing/selftests/riscv/vector/v_initval_nolibc.c new file +42
-0@@ -0,0 +1,68 @@ +// SPDX-License-Identifier: GPL-2.0-only + +#include "../../kselftest.h" +#define MAX_VSIZE (8192 * 32) + +void dump (char *ptr,int size) +{ +int i =0 ; + +for (i =0 ; i < size; i++) { +if (i !=0 ) { +if (i %16 ==0 ) +printf ("\n" ); +else if (i %8 ==0 ) +printf (" " ); + } +printf ("%02x " , ptr[i]); + } +printf ("\n" ); +} + +int main (void ) +{ +int i; +unsigned long vl; +char *datap, *tmp; + + datap =malloc (MAX_VSIZE ); +if (!datap) { +ksft_test_result_fail ("fail to allocate memory for size = %lu\n" ,MAX_VSIZE ); +exit (-1 ); + } + + tmp = datap; + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvli %0, x0, e8, m8, ta, ma\n\t" +"vse8.v v0, (%2)\n\t" +"add %1, %2, %0\n\t" +"vse8.v v8, (%1)\n\t" + "add %1, %1, %0\n\t"
tools/testing/selftests/riscv/vector/v_initval_nolibc.c line 52
Quoted 10 lines
tools/testing/selftests/riscv/vector/v_initval_nolibc.c new file +10
-0+"vse8.v v16, (%1)\n\t" +"add %1, %1, %0\n\t" +"vse8.v v24, (%1)\n\t" +".option pop\n\t" + : "=&r" (vl), "=r" (tmp) : "r" (datap) : "memory"); + + ksft_print_msg("vl = %lu\n", vl); + +if (datap[0 ] !=0x00 && datap[0 ] !=0xff ) { +ksft_test_result_fail ("v-regesters are not properly initialized\n" );
Nit: "v-registers"
tools/testing/selftests/riscv/vector/v_initval_nolibc.c line 59
Quoted 7 lines
tools/testing/selftests/riscv/vector/v_initval_nolibc.c new file +7
-0+dump (datap, vl *4 ); +exit (-1 ); + } + +for (i =1 ; i < vl *4 ; i++) { +if (datap[i] != datap[0 ]) { +ksft_test_result_fail ("detect stale values on v-regesters\n" );
Nit (dito): "v-registers", and maybe "detected".
With, or without the changes above,
Björn
_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Re: [v1, 2/2] selftests: Test RISC-V Vector's first-use handler
Quoted 50 lines · Björn Töpel
Quoted 47 lines · Andy Chiu
This add a test to check if the kernel zero-initializes all V registers after the first-use trap handler returns.
If V registers are not zero-initialized, then the test should fail one out of several runs:
``` root@sifive-fpga:~# ./v_initval_nolibc # vl = 256 not ok 1 detect stale values on v-regesters 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4c 41 4e 47 3d 43 0 50 41 54 48 3d 2f 75 73 72 2f 6c 6f 63 61 6c 2f 73 62 69 6e 3a 2f 75 73 72 2f 6c 6f 63 61 6c 2f 62 69 6e 3a 2f 75 73 72 ff ff 81 0 0 0 0 0 0 0 0 0 0 0 0 0 ```Otherwise, the test passes without errors each run.
Signed-off-by Andy Chiu <[email protected]>--- .../testing/selftests/riscv/vector/.gitignore | 1 + tools/testing/selftests/riscv/vector/Makefile | 6 +- .../selftests/riscv/vector/v_initval_nolibc.c | 68 +++++++++++++++++++ 3 files changed, 74 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/riscv/vector/v_initval_nolibc.ctools/testing/selftests/riscv/vector/.gitignore +1
-0@@ -1,2 +1,3 @@ vstate_exec_nolibc vstate_prctl +v_initval_nolibctools/testing/selftests/riscv/vector/Makefile +1
-1@@ -2,7 +2,7 @@ # Copyright (C) 2021 ARM Limited # Originally tools/testing/arm64/abi/Makefile -TEST_GEN_PROGS := vstate_prctl +TEST_GEN_PROGS := vstate_prctl v_initval_nolibc TEST_GEN_PROGS_EXTENDED := vstate_exec_nolibc
Quoted 12 lines
tools/testing/selftests/riscv/vector/Makefile $(OUTPUT)/vstate_prctl: vstate_prctl.c ../hwprobe/sys_hwprobe.S · line 19
Quoted 9 lines
tools/testing/selftests/riscv/vector/Makefile +4
-0include ../../lib.mk@@ -13,3 +13,7 @@ $(OUTPUT)/vstate_prctl: vstate_prctl.c ../hwprobe/sys_hwprobe.S $(OUTPUT)/vstate_exec_nolibc: vstate_exec_nolibc.c $(CC) -nostdlib -static -include ../../../../include/nolibc/nolibc.h \ -Wall $(CFLAGS) $(LDFLAGS) $^ -o $@ -lgcc + +$(OUTPUT)/v_initval_nolibc: v_initval_nolibc.c + $(CC) -nostdlib -static -include ../../../../include/nolibc/nolibc.h \ + -Wall $(CFLAGS) $(LDFLAGS) $^ -o $@ -lgccHmm, does this build with clang? (No, bigge on my end, and can be fixed later.)
Oops, I didn't try. I will respin the series if clang runs into problems. I am not very confident here, but do we need to gate build these tests if CONFIG_RISCV_ISA_V is not set? If my code fits clang well then it probably would be some toolchain dependency issues (guessing the ".option arch" part). Kconfig should be able to resolve this part.
Quoted 70 lines
Quoted 49 lines
tools/testing/selftests/riscv/vector/v_initval_nolibc.c new file +42
-0@@ -0,0 +1,68 @@ +// SPDX-License-Identifier: GPL-2.0-only + +#include "../../kselftest.h" +#define MAX_VSIZE (8192 * 32) + +void dump (char *ptr,int size) +{ +int i =0 ; + +for (i =0 ; i < size; i++) { +if (i !=0 ) { +if (i %16 ==0 ) +printf ("\n" ); +else if (i %8 ==0 ) +printf (" " ); + } +printf ("%02x " , ptr[i]); + } +printf ("\n" ); +} + +int main (void ) +{ +int i; +unsigned long vl; +char *datap, *tmp; + + datap =malloc (MAX_VSIZE ); +if (!datap) { +ksft_test_result_fail ("fail to allocate memory for size = %lu\n" ,MAX_VSIZE ); +exit (-1 ); + } + + tmp = datap; + asmvolatile ( +".option push\n\t" +".option arch, +v\n\t" +"vsetvli %0, x0, e8, m8, ta, ma\n\t" +"vse8.v v0, (%2)\n\t" +"add %1, %2, %0\n\t" +"vse8.v v8, (%1)\n\t" + "add %1, %1, %0\n\t"tools/testing/selftests/riscv/vector/v_initval_nolibc.c line 52
Quoted 10 lines
tools/testing/selftests/riscv/vector/v_initval_nolibc.c new file +10
-0+"vse8.v v16, (%1)\n\t" +"add %1, %1, %0\n\t" +"vse8.v v24, (%1)\n\t" +".option pop\n\t" + : "=&r" (vl), "=r" (tmp) : "r" (datap) : "memory"); + + ksft_print_msg("vl = %lu\n", vl); + +if (datap[0 ] !=0x00 && datap[0 ] !=0xff ) { +ksft_test_result_fail ("v-regesters are not properly initialized\n" );Nit: "v-registers"
tools/testing/selftests/riscv/vector/v_initval_nolibc.c line 59
Quoted 7 lines
tools/testing/selftests/riscv/vector/v_initval_nolibc.c new file +7
-0+dump (datap, vl *4 ); +exit (-1 ); + } + +for (i =1 ; i < vl *4 ; i++) { +if (datap[i] != datap[0 ]) { +ksft_test_result_fail ("detect stale values on v-regesters\n" );
Quoted 7 lines
Nit (dito): "v-registers", and maybe "detected".
With, or without the changes above,
Reviewed-by Björn Töpel <[email protected]>Björn
Thanks,
Andy
_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Palmer Dabbelt
AppliedQuoted 8 lines · Andy Chiu
Before applying this series, We only initialize the space for saving Vector registers. This is not enough as Vector registers themselves also neeeded to be initialized before dropping back into userspace. Or, we may risk leaking a process's data left in Vector register to another process. This can be verified by only applying the second patch and running the test.
[...]
Applied, thanks!
[1/2] riscv: vector: clear V-reg in the first-use trap
https://git.kernel.org/palmer/c/75b59f2a90aa [2/2] selftests: Test RISC-V Vector's first-use handler https://git.kernel.org/palmer/c/5c93c4c72fbc
Best regards,
Signature · 7 lines
Palmer Dabbelt <[email protected]> _______________________________________________ linux-riscv mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-riscv
This series was applied to riscv/linux.git (for-next) by Palmer Dabbelt <[email protected]>:
Hello:
This series was applied to riscv/linux.git (for-next) by Palmer Dabbelt <[email protected]>:
Quoted 8 lines · On Tue, 27 Jun 2023 01:55:53 +0000 you wrote:
Before applying this series, We only initialize the space for saving Vector registers. This is not enough as Vector registers themselves also neeeded to be initialized before dropping back into userspace. Or, we may risk leaking a process's data left in Vector register to another process. This can be verified by only applying the second patch and running the test.
[...]
Here is the summary with links:
- [v1,1/2] riscv: vector: clear V-reg in the first-use trap
https://git.kernel.org/riscv/c/75b59f2a90aa - [v1,2/2] selftests: Test RISC-V Vector's first-use handler https://git.kernel.org/riscv/c/5c93c4c72fbc
You are awesome, thank you!
Signature · 9 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html _______________________________________________ linux-riscv mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-riscv