diff mbox series

gcc-sanitizers: Add riscv64 as compatible host

Message ID 20241202043514.3046780-1-Deepthi.Hemraj@windriver.com
State New
Headers show
Series gcc-sanitizers: Add riscv64 as compatible host | expand

Commit Message

Hemraj, Deepthi Dec. 2, 2024, 4:35 a.m. UTC 
From: Deepthi Hemraj <Deepthi.Hemraj@windriver.com>

RISC-V offers several virtual memory address schemes (Sv39, Sv48, and Sv57),
but ASan currently supports only Sv39 on RISC-V64.
For RISC-V64 Sv39, ASan uses custom allocator configurations
tuned to manage large allocations efficiently.
These tunings are incompatible with larger address spaces like
Sv48/Sv57 due to differences in region sizes and alignment.

For riscv64, Asan's tuning for Sv39 can be enabled in qemu
by using the appropriate flag in the command line as shown below:
runqemu nographic qemuparams="-cpu rv64,sv39=true"

Additionally, the COMPATIBLE_HOST list in gcc-sanitizers has been
updated to include riscv64. All necessary tests were successfully
conducted on both hardware(Microchip PolarFire SoC)
and the qemurisv64 environment, with ASan effectively
detecting memory errors in both scenarios.

Signed-off-by: Deepthi Hemraj <Deepthi.Hemraj@windriver.com>
---
 meta/recipes-devtools/gcc/gcc-sanitizers.inc | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Comments

Randy MacLeod Dec. 2, 2024, 11:14 p.m. UTC | #1 
On 2024-12-01 11:35 p.m., Deepthi.Hemraj@windriver.com wrote:
> From: Deepthi Hemraj<Deepthi.Hemraj@windriver.com>
>
> RISC-V offers several virtual memory address schemes (Sv39, Sv48, and Sv57),
> but ASan currently supports only Sv39 on RISC-V64.
> For RISC-V64 Sv39, ASan uses custom allocator configurations
> tuned to manage large allocations efficiently.
> These tunings are incompatible with larger address spaces like
> Sv48/Sv57 due to differences in region sizes and alignment.
>
> For riscv64, Asan's tuning for Sv39 can be enabled in qemu
> by using the appropriate flag in the command line as shown below:
> runqemu nographic qemuparams="-cpu rv64,sv39=true"
>
> Additionally, the COMPATIBLE_HOST list in gcc-sanitizers has been
> updated to include riscv64. All necessary tests were successfully
> conducted on both hardware(Microchip PolarFire SoC)
> and the qemurisv64 environment, with ASan effectively
> detecting memory errors in both scenarios.


Hi Deepthi,


Nice! Thanks for confirming that the qemu command line option allows 
gcc-sanitizers to

work without any code changes.


One concern is testing: testers would have know that this flag was 
needed for gcc-sanitizers.

You could add a comment or documentation note as a hint but that doesn't 
automate and is error prone for people.

I did note that "spacemonkeydelivers" (!) maintains that :

"Indeed ASAN was ported to RISC-V when Sv39 was the default addressing 
mode, and changes to run it in Sv48/Sv57 must be made.
The issue is related to the way how the address space is split between 
stack, heap and shadow versions of them."
https://github.com/google/sanitizers/issues/1707#issuecomment-1852404560

so I don't think we've gotten to the best possible general purpose fix 
but rather a (helpful) work-around.

I don't think we should merge this since there's more work to do for a 
proper fix.

../Randy



>
> Signed-off-by: Deepthi Hemraj<Deepthi.Hemraj@windriver.com>
> ---
>   meta/recipes-devtools/gcc/gcc-sanitizers.inc | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/meta/recipes-devtools/gcc/gcc-sanitizers.inc b/meta/recipes-devtools/gcc/gcc-sanitizers.inc
> index 524ebd4ba4..8c98e9cf8a 100644
> --- a/meta/recipes-devtools/gcc/gcc-sanitizers.inc
> +++ b/meta/recipes-devtools/gcc/gcc-sanitizers.inc
> @@ -73,13 +73,14 @@ RRECOMMENDS:${PN}:append:x86 = " liblsan"
>   RRECOMMENDS:${PN}:append:x86-64 = " liblsan libtsan"
>   RRECOMMENDS:${PN}:append:powerpc64 = " liblsan libtsan"
>   RRECOMMENDS:${PN}:append:aarch64 = " liblsan libtsan"
> +RRECOMMENDS:${PN}:append:riscv64 = " liblsan libtsan"
>   
>   do_package_write_ipk[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>   do_package_write_deb[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>   do_package_write_rpm[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>   
>   # Only x86, powerpc, sparc, s390, arm, aarch64 and loongarch64 are supported
> -COMPATIBLE_HOST = '(x86_64|i.86|powerpc|sparc|s390|arm|aarch64|loongarch64).*-linux'
> +COMPATIBLE_HOST = '(x86_64|i.86|powerpc|sparc|s390|arm|aarch64|loongarch64|riscv64).*-linux'
>   # musl is currently broken entirely
>   COMPATIBLE_HOST:libc-musl = 'null'
>
Khem Raj Dec. 2, 2024, 11:50 p.m. UTC | #2 
On Mon, Dec 2, 2024 at 3:14 PM Randy MacLeod via
lists.openembedded.org
<randy.macleod=windriver.com@lists.openembedded.org> wrote:
>
> On 2024-12-01 11:35 p.m., Deepthi.Hemraj@windriver.com wrote:
>
> From: Deepthi Hemraj <Deepthi.Hemraj@windriver.com>
>
> RISC-V offers several virtual memory address schemes (Sv39, Sv48, and Sv57),
> but ASan currently supports only Sv39 on RISC-V64.
> For RISC-V64 Sv39, ASan uses custom allocator configurations
> tuned to manage large allocations efficiently.
> These tunings are incompatible with larger address spaces like
> Sv48/Sv57 due to differences in region sizes and alignment.
>
> For riscv64, Asan's tuning for Sv39 can be enabled in qemu
> by using the appropriate flag in the command line as shown below:
> runqemu nographic qemuparams="-cpu rv64,sv39=true"
>
> Additionally, the COMPATIBLE_HOST list in gcc-sanitizers has been
> updated to include riscv64. All necessary tests were successfully
> conducted on both hardware(Microchip PolarFire SoC)
> and the qemurisv64 environment, with ASan effectively
> detecting memory errors in both scenarios.
>
>
> Hi Deepthi,
>
>
> Nice! Thanks for confirming that the qemu command line option allows gcc-sanitizers to
>
> work without any code changes.
>
>
> One concern is testing: testers would have know that this flag was needed for gcc-sanitizers.
>
> You could add a comment or documentation note as a hint but that doesn't automate and is error prone for people.
>
> I did note that "spacemonkeydelivers" (!) maintains that :
>
> "Indeed ASAN was ported to RISC-V when Sv39 was the default addressing mode, and changes to run it in Sv48/Sv57 must be made.
> The issue is related to the way how the address space is split between stack, heap and shadow versions of them."
> https://github.com/google/sanitizers/issues/1707#issuecomment-1852404560

its ok to use Sv39, other modes are not as common yet.

>
> so I don't think we've gotten to the best possible general purpose fix but rather a (helpful) work-around.
>
> I don't think we should merge this since there's more work to do for a proper fix.
>
> ../Randy
>
>
>
>
> Signed-off-by: Deepthi Hemraj <Deepthi.Hemraj@windriver.com>
> ---
>  meta/recipes-devtools/gcc/gcc-sanitizers.inc | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/meta/recipes-devtools/gcc/gcc-sanitizers.inc b/meta/recipes-devtools/gcc/gcc-sanitizers.inc
> index 524ebd4ba4..8c98e9cf8a 100644
> --- a/meta/recipes-devtools/gcc/gcc-sanitizers.inc
> +++ b/meta/recipes-devtools/gcc/gcc-sanitizers.inc
> @@ -73,13 +73,14 @@ RRECOMMENDS:${PN}:append:x86 = " liblsan"
>  RRECOMMENDS:${PN}:append:x86-64 = " liblsan libtsan"
>  RRECOMMENDS:${PN}:append:powerpc64 = " liblsan libtsan"
>  RRECOMMENDS:${PN}:append:aarch64 = " liblsan libtsan"
> +RRECOMMENDS:${PN}:append:riscv64 = " liblsan libtsan"
>
>  do_package_write_ipk[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>  do_package_write_deb[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>  do_package_write_rpm[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>
>  # Only x86, powerpc, sparc, s390, arm, aarch64 and loongarch64 are supported
> -COMPATIBLE_HOST = '(x86_64|i.86|powerpc|sparc|s390|arm|aarch64|loongarch64).*-linux'
> +COMPATIBLE_HOST = '(x86_64|i.86|powerpc|sparc|s390|arm|aarch64|loongarch64|riscv64).*-linux'
>  # musl is currently broken entirely
>  COMPATIBLE_HOST:libc-musl = 'null'
>
>
>
> --
> # Randy MacLeod
> # Wind River Linux
>
>
> -=-=-=-=-=-=-=-=-=-=-=-
> Links: You receive all messages sent to this group.
> View/Reply Online (#208161): https://lists.openembedded.org/g/openembedded-core/message/208161
> Mute This Topic: https://lists.openembedded.org/mt/109874858/1997914
> Group Owner: openembedded-core+owner@lists.openembedded.org
> Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [raj.khem@gmail.com]
> -=-=-=-=-=-=-=-=-=-=-=-
>
Randy MacLeod Dec. 3, 2024, 1:14 a.m. UTC | #3 
On 2024-12-02 6:50 p.m., Khem Raj wrote:
> On Mon, Dec 2, 2024 at 3:14 PM Randy MacLeod via
> lists.openembedded.org
> <randy.macleod=windriver.com@lists.openembedded.org> wrote:
>> On 2024-12-01 11:35 p.m.,Deepthi.Hemraj@windriver.com wrote:
>>
>> From: Deepthi Hemraj<Deepthi.Hemraj@windriver.com>
>>
>> RISC-V offers several virtual memory address schemes (Sv39, Sv48, and Sv57),
>> but ASan currently supports only Sv39 on RISC-V64.
>> For RISC-V64 Sv39, ASan uses custom allocator configurations
>> tuned to manage large allocations efficiently.
>> These tunings are incompatible with larger address spaces like
>> Sv48/Sv57 due to differences in region sizes and alignment.
>>
>> For riscv64, Asan's tuning for Sv39 can be enabled in qemu
>> by using the appropriate flag in the command line as shown below:
>> runqemu nographic qemuparams="-cpu rv64,sv39=true"
>>
>> Additionally, the COMPATIBLE_HOST list in gcc-sanitizers has been
>> updated to include riscv64. All necessary tests were successfully
>> conducted on both hardware(Microchip PolarFire SoC)
>> and the qemurisv64 environment, with ASan effectively
>> detecting memory errors in both scenarios.
>>
>>
>> Hi Deepthi,
>>
>>
>> Nice! Thanks for confirming that the qemu command line option allows gcc-sanitizers to
>>
>> work without any code changes.
>>
>>
>> One concern is testing: testers would have know that this flag was needed for gcc-sanitizers.
>>
>> You could add a comment or documentation note as a hint but that doesn't automate and is error prone for people.
>>
>> I did note that "spacemonkeydelivers" (!) maintains that :
>>
>> "Indeed ASAN was ported to RISC-V when Sv39 was the default addressing mode, and changes to run it in Sv48/Sv57 must be made.
>> The issue is related to the way how the address space is split between stack, heap and shadow versions of them."
>> https://github.com/google/sanitizers/issues/1707#issuecomment-1852404560
> its ok to use Sv39, other modes are not as common yet.

Khem,

Are you saying that we should config qemu-native (target?) for riscv64 
to default to SV39 ?

../Randy

>
>> so I don't think we've gotten to the best possible general purpose fix but rather a (helpful) work-around.
>>
>> I don't think we should merge this since there's more work to do for a proper fix.
>>
>> ../Randy
>>
>>
>>
>>
>> Signed-off-by: Deepthi Hemraj<Deepthi.Hemraj@windriver.com>
>> ---
>>   meta/recipes-devtools/gcc/gcc-sanitizers.inc | 3 ++-
>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/meta/recipes-devtools/gcc/gcc-sanitizers.inc b/meta/recipes-devtools/gcc/gcc-sanitizers.inc
>> index 524ebd4ba4..8c98e9cf8a 100644
>> --- a/meta/recipes-devtools/gcc/gcc-sanitizers.inc
>> +++ b/meta/recipes-devtools/gcc/gcc-sanitizers.inc
>> @@ -73,13 +73,14 @@ RRECOMMENDS:${PN}:append:x86 = " liblsan"
>>   RRECOMMENDS:${PN}:append:x86-64 = " liblsan libtsan"
>>   RRECOMMENDS:${PN}:append:powerpc64 = " liblsan libtsan"
>>   RRECOMMENDS:${PN}:append:aarch64 = " liblsan libtsan"
>> +RRECOMMENDS:${PN}:append:riscv64 = " liblsan libtsan"
>>
>>   do_package_write_ipk[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>>   do_package_write_deb[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>>   do_package_write_rpm[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>>
>>   # Only x86, powerpc, sparc, s390, arm, aarch64 and loongarch64 are supported
>> -COMPATIBLE_HOST = '(x86_64|i.86|powerpc|sparc|s390|arm|aarch64|loongarch64).*-linux'
>> +COMPATIBLE_HOST = '(x86_64|i.86|powerpc|sparc|s390|arm|aarch64|loongarch64|riscv64).*-linux'
>>   # musl is currently broken entirely
>>   COMPATIBLE_HOST:libc-musl = 'null'
>>
>>
>>
>> --
>> # Randy MacLeod
>> # Wind River Linux
>>
>>
>> -=-=-=-=-=-=-=-=-=-=-=-
>> Links: You receive all messages sent to this group.
>> View/Reply Online (#208161):https://lists.openembedded.org/g/openembedded-core/message/208161
>> Mute This Topic:https://lists.openembedded.org/mt/109874858/1997914
>> Group Owner:openembedded-core+owner@lists.openembedded.org
>> Unsubscribe:https://lists.openembedded.org/g/openembedded-core/unsub [raj.khem@gmail.com]
>> -=-=-=-=-=-=-=-=-=-=-=-
>>
Khem Raj Dec. 3, 2024, 1:54 a.m. UTC | #4 
On Mon, Dec 2, 2024 at 5:14 PM Randy MacLeod
<randy.macleod@windriver.com> wrote:
>
> On 2024-12-02 6:50 p.m., Khem Raj wrote:
>
> On Mon, Dec 2, 2024 at 3:14 PM Randy MacLeod via
> lists.openembedded.org
> <randy.macleod=windriver.com@lists.openembedded.org> wrote:
>
> On 2024-12-01 11:35 p.m., Deepthi.Hemraj@windriver.com wrote:
>
> From: Deepthi Hemraj <Deepthi.Hemraj@windriver.com>
>
> RISC-V offers several virtual memory address schemes (Sv39, Sv48, and Sv57),
> but ASan currently supports only Sv39 on RISC-V64.
> For RISC-V64 Sv39, ASan uses custom allocator configurations
> tuned to manage large allocations efficiently.
> These tunings are incompatible with larger address spaces like
> Sv48/Sv57 due to differences in region sizes and alignment.
>
> For riscv64, Asan's tuning for Sv39 can be enabled in qemu
> by using the appropriate flag in the command line as shown below:
> runqemu nographic qemuparams="-cpu rv64,sv39=true"
>
> Additionally, the COMPATIBLE_HOST list in gcc-sanitizers has been
> updated to include riscv64. All necessary tests were successfully
> conducted on both hardware(Microchip PolarFire SoC)
> and the qemurisv64 environment, with ASan effectively
> detecting memory errors in both scenarios.
>
>
> Hi Deepthi,
>
>
> Nice! Thanks for confirming that the qemu command line option allows gcc-sanitizers to
>
> work without any code changes.
>
>
> One concern is testing: testers would have know that this flag was needed for gcc-sanitizers.
>
> You could add a comment or documentation note as a hint but that doesn't automate and is error prone for people.
>
> I did note that "spacemonkeydelivers" (!) maintains that :
>
> "Indeed ASAN was ported to RISC-V when Sv39 was the default addressing mode, and changes to run it in Sv48/Sv57 must be made.
> The issue is related to the way how the address space is split between stack, heap and shadow versions of them."
> https://github.com/google/sanitizers/issues/1707#issuecomment-1852404560
>
> its ok to use Sv39, other modes are not as common yet.
>
> Khem,
>
> Are you saying that we should config qemu-native (target?) for riscv64 to default to SV39 ?

yes with 6.6+ kernel it should work ok.

>
> ../Randy
>
> so I don't think we've gotten to the best possible general purpose fix but rather a (helpful) work-around.
>
> I don't think we should merge this since there's more work to do for a proper fix.
>
> ../Randy
>
>
>
>
> Signed-off-by: Deepthi Hemraj <Deepthi.Hemraj@windriver.com>
> ---
>  meta/recipes-devtools/gcc/gcc-sanitizers.inc | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/meta/recipes-devtools/gcc/gcc-sanitizers.inc b/meta/recipes-devtools/gcc/gcc-sanitizers.inc
> index 524ebd4ba4..8c98e9cf8a 100644
> --- a/meta/recipes-devtools/gcc/gcc-sanitizers.inc
> +++ b/meta/recipes-devtools/gcc/gcc-sanitizers.inc
> @@ -73,13 +73,14 @@ RRECOMMENDS:${PN}:append:x86 = " liblsan"
>  RRECOMMENDS:${PN}:append:x86-64 = " liblsan libtsan"
>  RRECOMMENDS:${PN}:append:powerpc64 = " liblsan libtsan"
>  RRECOMMENDS:${PN}:append:aarch64 = " liblsan libtsan"
> +RRECOMMENDS:${PN}:append:riscv64 = " liblsan libtsan"
>
>  do_package_write_ipk[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>  do_package_write_deb[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>  do_package_write_rpm[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
>
>  # Only x86, powerpc, sparc, s390, arm, aarch64 and loongarch64 are supported
> -COMPATIBLE_HOST = '(x86_64|i.86|powerpc|sparc|s390|arm|aarch64|loongarch64).*-linux'
> +COMPATIBLE_HOST = '(x86_64|i.86|powerpc|sparc|s390|arm|aarch64|loongarch64|riscv64).*-linux'
>  # musl is currently broken entirely
>  COMPATIBLE_HOST:libc-musl = 'null'
>
>
>
> --
> # Randy MacLeod
> # Wind River Linux
>
>
> -=-=-=-=-=-=-=-=-=-=-=-
> Links: You receive all messages sent to this group.
> View/Reply Online (#208161): https://lists.openembedded.org/g/openembedded-core/message/208161
> Mute This Topic: https://lists.openembedded.org/mt/109874858/1997914
> Group Owner: openembedded-core+owner@lists.openembedded.org
> Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [raj.khem@gmail.com]
> -=-=-=-=-=-=-=-=-=-=-=-
>
>
> --
> # Randy MacLeod
> # Wind River Linux
diff mbox series

Patch

diff --git a/meta/recipes-devtools/gcc/gcc-sanitizers.inc b/meta/recipes-devtools/gcc/gcc-sanitizers.inc
index 524ebd4ba4..8c98e9cf8a 100644
--- a/meta/recipes-devtools/gcc/gcc-sanitizers.inc
+++ b/meta/recipes-devtools/gcc/gcc-sanitizers.inc
@@ -73,13 +73,14 @@  RRECOMMENDS:${PN}:append:x86 = " liblsan"
 RRECOMMENDS:${PN}:append:x86-64 = " liblsan libtsan"
 RRECOMMENDS:${PN}:append:powerpc64 = " liblsan libtsan"
 RRECOMMENDS:${PN}:append:aarch64 = " liblsan libtsan"
+RRECOMMENDS:${PN}:append:riscv64 = " liblsan libtsan"
 
 do_package_write_ipk[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
 do_package_write_deb[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
 do_package_write_rpm[depends] += "virtual/${MLPREFIX}${TARGET_PREFIX}compilerlibs:do_packagedata"
 
 # Only x86, powerpc, sparc, s390, arm, aarch64 and loongarch64 are supported
-COMPATIBLE_HOST = '(x86_64|i.86|powerpc|sparc|s390|arm|aarch64|loongarch64).*-linux'
+COMPATIBLE_HOST = '(x86_64|i.86|powerpc|sparc|s390|arm|aarch64|loongarch64|riscv64).*-linux'
 # musl is currently broken entirely
 COMPATIBLE_HOST:libc-musl = 'null'