Improve the performance of icache flushing by creating a new prctl flag
PR_RISCV_SET_ICACHE_FLUSH_CTX. The interface is left generic to allow
for future expansions such as with the proposed J extension [1].
Documentation is also provided to explain the use case.
[1] https://github.com/riscv/riscv-j-extension
Signed-off-by: Charlie Jenkins <[email protected]>
---
Changes in v3:
- Check if value force_icache_flush set on thread, rather than in mm
twice (Clément)
- Link to v2: https://lore.kernel.org/r/[email protected]
Changes in v2:
- Fix kernel-doc comment (Conor)
- Link to v1: https://lore.kernel.org/r/[email protected]
---
Charlie Jenkins (2):
riscv: Include riscv_set_icache_flush_ctx prctl
documentation: Document PR_RISCV_SET_ICACHE_FLUSH_CTX prctl
Documentation/arch/riscv/cmodx.rst | 98 ++++++++++++++++++++++++++++++++++++++
Documentation/arch/riscv/index.rst | 1 +
arch/riscv/include/asm/mmu.h | 2 +
arch/riscv/include/asm/processor.h | 6 +++
arch/riscv/mm/cacheflush.c | 37 ++++++++++++++
arch/riscv/mm/context.c | 8 ++--
include/uapi/linux/prctl.h | 3 ++
kernel/sys.c | 6 +++
8 files changed, 158 insertions(+), 3 deletions(-)
---
base-commit: b85ea95d086471afb4ad062012a4d73cd328fa86
change-id: 20231117-fencei-f9f60d784fa0
--
- Charlie
Support new prctl with key PR_RISCV_SET_ICACHE_FLUSH_CTX to enable
optimization of cross modifying code. This prctl enables userspace code
to use icache flushing instructions such as fence.i with the guarantee
that the icache will continue to be clean after thread migration.
Signed-off-by: Charlie Jenkins <[email protected]>
---
arch/riscv/include/asm/mmu.h | 2 ++
arch/riscv/include/asm/processor.h | 6 ++++++
arch/riscv/mm/cacheflush.c | 37 +++++++++++++++++++++++++++++++++++++
arch/riscv/mm/context.c | 8 +++++---
include/uapi/linux/prctl.h | 3 +++
kernel/sys.c | 6 ++++++
6 files changed, 59 insertions(+), 3 deletions(-)
diff --git a/arch/riscv/include/asm/mmu.h b/arch/riscv/include/asm/mmu.h
index 355504b37f8e..60be458e94da 100644
--- a/arch/riscv/include/asm/mmu.h
+++ b/arch/riscv/include/asm/mmu.h
@@ -19,6 +19,8 @@ typedef struct {
#ifdef CONFIG_SMP
/* A local icache flush is needed before user execution can resume. */
cpumask_t icache_stale_mask;
+ /* Force local icache flush on all migrations. */
+ bool force_icache_flush;
#endif
#ifdef CONFIG_BINFMT_ELF_FDPIC
unsigned long exec_fdpic_loadmap;
diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
index f19f861cda54..7eda6c75e0f2 100644
--- a/arch/riscv/include/asm/processor.h
+++ b/arch/riscv/include/asm/processor.h
@@ -84,6 +84,9 @@ struct thread_struct {
unsigned long vstate_ctrl;
struct __riscv_v_ext_state vstate;
unsigned long align_ctl;
+#ifdef CONFIG_SMP
+ bool force_icache_flush;
+#endif
};
/* Whitelist the fstate from the task_struct for hardened usercopy */
@@ -145,6 +148,9 @@ extern int set_unalign_ctl(struct task_struct *tsk, unsigned int val);
#define GET_UNALIGN_CTL(tsk, addr) get_unalign_ctl((tsk), (addr))
#define SET_UNALIGN_CTL(tsk, val) set_unalign_ctl((tsk), (val))
+#define RISCV_SET_ICACHE_FLUSH_CTX(arg1, arg2) riscv_set_icache_flush_ctx(arg1, arg2)
+extern int riscv_set_icache_flush_ctx(unsigned long ctx, unsigned long per_thread);
+
#endif /* __ASSEMBLY__ */
#endif /* _ASM_RISCV_PROCESSOR_H */
diff --git a/arch/riscv/mm/cacheflush.c b/arch/riscv/mm/cacheflush.c
index 55a34f2020a8..3b2bf8256a10 100644
--- a/arch/riscv/mm/cacheflush.c
+++ b/arch/riscv/mm/cacheflush.c
@@ -5,6 +5,7 @@
#include <linux/acpi.h>
#include <linux/of.h>
+#include <linux/prctl.h>
#include <asm/acpi.h>
#include <asm/cacheflush.h>
@@ -152,3 +153,39 @@ void __init riscv_init_cbo_blocksizes(void)
if (cboz_block_size)
riscv_cboz_block_size = cboz_block_size;
}
+
+/**
+ * riscv_set_icache_flush_ctx() - Enable userspace to emit icache flushing instructions.
+ * @ctx: Sets the type of context
+ * - PR_RISCV_CTX_SW_FENCEI: Allow fence.i in userspace. Another fence.i will
+ * emitted on thread/process migration.
+ * @per_thread: When set to 0, will use the default behavior of setting the
+ * icache flush context per process. When set to 1, will use a per thread
+ * context.
+ *
+ * When in per-process context, there may be multiple threads using the same mm.
+ * Therefore, the icache can never be assumed clean when. Multiple threads in
+ * the process may modify instructions in the mm concurrently.
+ *
+ * In per-thread context, it can be assumed that all modifications to
+ * instructions in memory will be performed by this thread. When the thread is
+ * migrated the icache will be flushed.
+ *
+ */
+int riscv_set_icache_flush_ctx(unsigned long ctx, unsigned long per_thread)
+{
+#ifdef CONFIG_SMP
+ switch (ctx) {
+ case PR_RISCV_CTX_SW_FENCEI:
+ if (per_thread)
+ current->thread.force_icache_flush = true;
+ else
+ current->mm->context.force_icache_flush = true;
+ break;
+
+ default:
+ break;
+ }
+#endif
+ return 0;
+}
diff --git a/arch/riscv/mm/context.c b/arch/riscv/mm/context.c
index 217fd4de6134..0146c61be0ab 100644
--- a/arch/riscv/mm/context.c
+++ b/arch/riscv/mm/context.c
@@ -297,12 +297,14 @@ static inline void set_mm(struct mm_struct *prev,
*
* The "cpu" argument must be the current local CPU number.
*/
-static inline void flush_icache_deferred(struct mm_struct *mm, unsigned int cpu)
+static inline void flush_icache_deferred(struct mm_struct *mm, unsigned int cpu,
+ struct task_struct *task)
{
#ifdef CONFIG_SMP
cpumask_t *mask = &mm->context.icache_stale_mask;
- if (cpumask_test_cpu(cpu, mask)) {
+ if (cpumask_test_cpu(cpu, mask) || mm->context.force_icache_flush ||
+ (task && task->thread.force_icache_flush)) {
cpumask_clear_cpu(cpu, mask);
/*
* Ensure the remote hart's writes are visible to this hart.
@@ -332,5 +334,5 @@ void switch_mm(struct mm_struct *prev, struct mm_struct *next,
set_mm(prev, next, cpu);
- flush_icache_deferred(next, cpu);
+ flush_icache_deferred(next, cpu, task);
}
diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h
index 370ed14b1ae0..472801ea78cc 100644
--- a/include/uapi/linux/prctl.h
+++ b/include/uapi/linux/prctl.h
@@ -306,4 +306,7 @@ struct prctl_mm_map {
# define PR_RISCV_V_VSTATE_CTRL_NEXT_MASK 0xc
# define PR_RISCV_V_VSTATE_CTRL_MASK 0x1f
+#define PR_RISCV_SET_ICACHE_FLUSH_CTX 71
+# define PR_RISCV_CTX_SW_FENCEI 0
+
#endif /* _LINUX_PRCTL_H */
diff --git a/kernel/sys.c b/kernel/sys.c
index 420d9cb9cc8e..e806a8a67c36 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -146,6 +146,9 @@
#ifndef RISCV_V_GET_CONTROL
# define RISCV_V_GET_CONTROL() (-EINVAL)
#endif
+#ifndef RISCV_SET_ICACHE_FLUSH_CTX
+# define RISCV_SET_ICACHE_FLUSH_CTX(a, b) (-EINVAL)
+#endif
/*
* this is where the system-wide overflow UID and GID are defined, for
@@ -2739,6 +2742,9 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
case PR_RISCV_V_GET_CONTROL:
error = RISCV_V_GET_CONTROL();
break;
+ case PR_RISCV_SET_ICACHE_FLUSH_CTX:
+ error = RISCV_SET_ICACHE_FLUSH_CTX(arg2, arg3);
+ break;
default:
error = -EINVAL;
break;
--
2.43.0
Provide documentation that explains how to properly do CMODX in riscv.
Signed-off-by: Charlie Jenkins <[email protected]>
---
Documentation/arch/riscv/cmodx.rst | 98 ++++++++++++++++++++++++++++++++++++++
Documentation/arch/riscv/index.rst | 1 +
2 files changed, 99 insertions(+)
diff --git a/Documentation/arch/riscv/cmodx.rst b/Documentation/arch/riscv/cmodx.rst
new file mode 100644
index 000000000000..20f327d85116
--- /dev/null
+++ b/Documentation/arch/riscv/cmodx.rst
@@ -0,0 +1,98 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+==============================================================================
+Concurrent Modification and Execution of Instructions (CMODX) for RISC-V Linux
+==============================================================================
+
+CMODX is a programming technique where a program executes instructions that were
+modified by the program itself. Instruction storage and the instruction cache
+(icache) is not guaranteed to be synchronized on RISC-V hardware. Therefore, the
+program must enforce its own synchonization with the unprivileged fence.i
+instruction.
+
+However, the default Linux ABI prohibits the use of fence.i in userspace
+applications. At any point the scheduler may migrate a task onto a new hart. If
+migration occurs after the userspace synchronized the icache and instruction
+storage with fence.i, the icache will no longer be clean. This is due to the
+behavior of fence.i only affecting the hart that it is called on. Thus, the hart
+that the task has been migrated to, may not have synchronized instruction
+storage and icache.
+
+There are two ways to solve this problem: use the riscv_flush_icache() syscall,
+or use the ``PR_RISCV_SET_ICACHE_FLUSH_CTX`` prctl(). The syscall should be used
+when the application very rarely needs to flush the icache. If the icache will
+need to be flushed many times in the lifetime of the application, the prctl
+should be used.
+
+The prctl informs the kernel that it must emit synchronizing instructions upon
+task migration. The program itself must emit synchonizing instructions when
+necessary as well.
+
+1. prctl() Interface
+---------------------
+
+Before the program emits their first icache flushing instruction, the program
+must call this prctl().
+
+* prctl(PR_RISCV_SET_ICACHE_FLUSH_CTX, unsigned long ctx, unsigned long per_thread)
+
+ Sets the icache flushing context. If per_thread is 0, context will be
+ applied per process, otherwise if per_thread is 1 context will be
+ per-thread. Any other number will have undefined behavior.
+
+ * :c:macro:`PR_RISCV_CTX_SW_FENCEI`: Allow fence.i to be called in
+ userspace.
+
+Example usage:
+
+The following files are meant to be compiled and linked with each other. The
+modify_instruction() function replaces an add with 0 with an add with one,
+causing the instruction sequence in get_value() to change from returning a zero
+to returning a one.
+
+cmodx.c::
+
+ #include <stdio.h>
+ #include <sys/prctl.h>
+
+ extern int get_value();
+ extern void modify_instruction();
+
+ int main()
+ {
+ int value = get_value();
+ printf("Value before cmodx: %d\n", value);
+
+ // Call prctl before first fence.i is called inside modify_instruction
+ prctl(PR_RISCV_SET_ICACHE_FLUSH_CTX, PR_RISCV_CTX_SW_FENCEI, 0);
+ modify_instruction();
+
+ value = get_value();
+ printf("Value after cmodx: %d\n", value);
+ return 0;
+ }
+
+cmodx.S::
+
+ .option norvc
+
+ .text
+ .global modify_instruction
+ modify_instruction:
+ lw a0, new_insn
+ lui a5,%hi(old_insn)
+ sw a0,%lo(old_insn)(a5)
+ fence.i
+ ret
+
+ .section modifiable, "awx"
+ .global get_value
+ get_value:
+ li a0, 0
+ old_insn:
+ addi a0, a0, 0
+ ret
+
+ .data
+ new_insn:
+ addi a0, a0, 1
diff --git a/Documentation/arch/riscv/index.rst b/Documentation/arch/riscv/index.rst
index 4dab0cb4b900..eecf347ce849 100644
--- a/Documentation/arch/riscv/index.rst
+++ b/Documentation/arch/riscv/index.rst
@@ -13,6 +13,7 @@ RISC-V architecture
patch-acceptance
uabi
vector
+ cmodx
features
--
2.43.0
On Wed, Dec 13, 2023 at 2:48 PM Charlie Jenkins <[email protected]> wrote:
>
> Provide documentation that explains how to properly do CMODX in riscv.
>
> Signed-off-by: Charlie Jenkins <[email protected]>
> ---
> Documentation/arch/riscv/cmodx.rst | 98 ++++++++++++++++++++++++++++++++++++++
> Documentation/arch/riscv/index.rst | 1 +
> 2 files changed, 99 insertions(+)
>
> diff --git a/Documentation/arch/riscv/cmodx.rst b/Documentation/arch/riscv/cmodx.rst
> new file mode 100644
> index 000000000000..20f327d85116
> --- /dev/null
> +++ b/Documentation/arch/riscv/cmodx.rst
> @@ -0,0 +1,98 @@
> +.. SPDX-License-Identifier: GPL-2.0
> +
> +==============================================================================
> +Concurrent Modification and Execution of Instructions (CMODX) for RISC-V Linux
> +==============================================================================
> +
> +CMODX is a programming technique where a program executes instructions that were
> +modified by the program itself. Instruction storage and the instruction cache
> +(icache) is not guaranteed to be synchronized on RISC-V hardware. Therefore, the
> +program must enforce its own synchonization with the unprivileged fence.i/
/s/synchonization/synchronization
> +instruction.
> +
> +However, the default Linux ABI prohibits the use of fence.i in userspace
> +applications. At any point the scheduler may migrate a task onto a new hart. If
> +migration occurs after the userspace synchronized the icache and instruction
> +storage with fence.i, the icache will no longer be clean. This is due to the
> +behavior of fence.i only affecting the hart that it is called on. Thus, the hart
> +that the task has been migrated to, may not have synchronized instruction
> +storage and icache.
> +
> +There are two ways to solve this problem: use the riscv_flush_icache() syscall,
> +or use the ``PR_RISCV_SET_ICACHE_FLUSH_CTX`` prctl(). The syscall should be used
> +when the application very rarely needs to flush the icache. If the icache will
The syscall is a one time operation while prctl is sticky.
It would be great if we can add a little more context why the syscall
behaves this way compared to prctl.
> +need to be flushed many times in the lifetime of the application, the prctl
> +should be used.
> +
> +The prctl informs the kernel that it must emit synchronizing instructions upon
> +task migration. The program itself must emit synchonizing instructions when
/s/synchonizing/synchronizing
> +necessary as well.
> +
> +1. prctl() Interface
> +---------------------
> +
> +Before the program emits their first icache flushing instruction, the program
> +must call this prctl().
> +
> +* prctl(PR_RISCV_SET_ICACHE_FLUSH_CTX, unsigned long ctx, unsigned long per_thread)
> +
> + Sets the icache flushing context. If per_thread is 0, context will be
> + applied per process, otherwise if per_thread is 1 context will be
> + per-thread. Any other number will have undefined behavior.
> +
> + * :c:macro:`PR_RISCV_CTX_SW_FENCEI`: Allow fence.i to be called in
> + userspace.
> +
> +Example usage:
> +
> +The following files are meant to be compiled and linked with each other. The
> +modify_instruction() function replaces an add with 0 with an add with one,
> +causing the instruction sequence in get_value() to change from returning a zero
> +to returning a one.
> +
> +cmodx.c::
> +
> + #include <stdio.h>
> + #include <sys/prctl.h>
> +
> + extern int get_value();
> + extern void modify_instruction();
> +
> + int main()
> + {
> + int value = get_value();
> + printf("Value before cmodx: %d\n", value);
> +
> + // Call prctl before first fence.i is called inside modify_instruction
> + prctl(PR_RISCV_SET_ICACHE_FLUSH_CTX, PR_RISCV_CTX_SW_FENCEI, 0);
> + modify_instruction();
> +
> + value = get_value();
> + printf("Value after cmodx: %d\n", value);
> + return 0;
> + }
> +
> +cmodx.S::
> +
> + .option norvc
> +
> + .text
> + .global modify_instruction
> + modify_instruction:
> + lw a0, new_insn
> + lui a5,%hi(old_insn)
> + sw a0,%lo(old_insn)(a5)
> + fence.i
> + ret
> +
> + .section modifiable, "awx"
> + .global get_value
> + get_value:
> + li a0, 0
> + old_insn:
> + addi a0, a0, 0
> + ret
> +
> + .data
> + new_insn:
> + addi a0, a0, 1
> diff --git a/Documentation/arch/riscv/index.rst b/Documentation/arch/riscv/index.rst
> index 4dab0cb4b900..eecf347ce849 100644
> --- a/Documentation/arch/riscv/index.rst
> +++ b/Documentation/arch/riscv/index.rst
> @@ -13,6 +13,7 @@ RISC-V architecture
> patch-acceptance
> uabi
> vector
> + cmodx
>
> features
>
>
> --
> 2.43.0
>
>
> _______________________________________________
> linux-riscv mailing list
> [email protected]
> http://lists.infradead.org/mailman/listinfo/linux-riscv
--
Regards,
Atish
On Wed, Dec 13, 2023 at 2:48 PM Charlie Jenkins <[email protected]> wrote:
>
> Support new prctl with key PR_RISCV_SET_ICACHE_FLUSH_CTX to enable
> optimization of cross modifying code. This prctl enables userspace code
> to use icache flushing instructions such as fence.i with the guarantee
> that the icache will continue to be clean after thread migration.
>
Do we need a CLEAR option as well ? I am wondering if there is a use case where
an application wants the prctl for only certain duration but may
invoke fence.i multiple times (syscall will still not be a good option
for them.)
> Signed-off-by: Charlie Jenkins <[email protected]>
> ---
> arch/riscv/include/asm/mmu.h | 2 ++
> arch/riscv/include/asm/processor.h | 6 ++++++
> arch/riscv/mm/cacheflush.c | 37 +++++++++++++++++++++++++++++++++++++
> arch/riscv/mm/context.c | 8 +++++---
> include/uapi/linux/prctl.h | 3 +++
> kernel/sys.c | 6 ++++++
> 6 files changed, 59 insertions(+), 3 deletions(-)
>
> diff --git a/arch/riscv/include/asm/mmu.h b/arch/riscv/include/asm/mmu.h
> index 355504b37f8e..60be458e94da 100644
> --- a/arch/riscv/include/asm/mmu.h
> +++ b/arch/riscv/include/asm/mmu.h
> @@ -19,6 +19,8 @@ typedef struct {
> #ifdef CONFIG_SMP
> /* A local icache flush is needed before user execution can resume. */
> cpumask_t icache_stale_mask;
> + /* Force local icache flush on all migrations. */
> + bool force_icache_flush;
> #endif
> #ifdef CONFIG_BINFMT_ELF_FDPIC
> unsigned long exec_fdpic_loadmap;
> diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
> index f19f861cda54..7eda6c75e0f2 100644
> --- a/arch/riscv/include/asm/processor.h
> +++ b/arch/riscv/include/asm/processor.h
> @@ -84,6 +84,9 @@ struct thread_struct {
> unsigned long vstate_ctrl;
> struct __riscv_v_ext_state vstate;
> unsigned long align_ctl;
> +#ifdef CONFIG_SMP
> + bool force_icache_flush;
> +#endif
> };
>
> /* Whitelist the fstate from the task_struct for hardened usercopy */
> @@ -145,6 +148,9 @@ extern int set_unalign_ctl(struct task_struct *tsk, unsigned int val);
> #define GET_UNALIGN_CTL(tsk, addr) get_unalign_ctl((tsk), (addr))
> #define SET_UNALIGN_CTL(tsk, val) set_unalign_ctl((tsk), (val))
>
> +#define RISCV_SET_ICACHE_FLUSH_CTX(arg1, arg2) riscv_set_icache_flush_ctx(arg1, arg2)
> +extern int riscv_set_icache_flush_ctx(unsigned long ctx, unsigned long per_thread);
> +
> #endif /* __ASSEMBLY__ */
>
> #endif /* _ASM_RISCV_PROCESSOR_H */
> diff --git a/arch/riscv/mm/cacheflush.c b/arch/riscv/mm/cacheflush.c
> index 55a34f2020a8..3b2bf8256a10 100644
> --- a/arch/riscv/mm/cacheflush.c
> +++ b/arch/riscv/mm/cacheflush.c
> @@ -5,6 +5,7 @@
>
> #include <linux/acpi.h>
> #include <linux/of.h>
> +#include <linux/prctl.h>
> #include <asm/acpi.h>
> #include <asm/cacheflush.h>
>
> @@ -152,3 +153,39 @@ void __init riscv_init_cbo_blocksizes(void)
> if (cboz_block_size)
> riscv_cboz_block_size = cboz_block_size;
> }
> +
> +/**
> + * riscv_set_icache_flush_ctx() - Enable userspace to emit icache flushing instructions.
> + * @ctx: Sets the type of context
> + * - PR_RISCV_CTX_SW_FENCEI: Allow fence.i in userspace. Another fence.i will
> + * emitted on thread/process migration.
> + * @per_thread: When set to 0, will use the default behavior of setting the
> + * icache flush context per process. When set to 1, will use a per thread
> + * context.
> + *
> + * When in per-process context, there may be multiple threads using the same mm.
> + * Therefore, the icache can never be assumed clean when. Multiple threads in
> + * the process may modify instructions in the mm concurrently.
> + *
> + * In per-thread context, it can be assumed that all modifications to
> + * instructions in memory will be performed by this thread. When the thread is
> + * migrated the icache will be flushed.
> + *
> + */
> +int riscv_set_icache_flush_ctx(unsigned long ctx, unsigned long per_thread)
> +{
> +#ifdef CONFIG_SMP
> + switch (ctx) {
> + case PR_RISCV_CTX_SW_FENCEI:
> + if (per_thread)
> + current->thread.force_icache_flush = true;
> + else
> + current->mm->context.force_icache_flush = true;
> + break;
> +
> + default:
> + break;
> + }
> +#endif
> + return 0;
> +}
> diff --git a/arch/riscv/mm/context.c b/arch/riscv/mm/context.c
> index 217fd4de6134..0146c61be0ab 100644
> --- a/arch/riscv/mm/context.c
> +++ b/arch/riscv/mm/context.c
> @@ -297,12 +297,14 @@ static inline void set_mm(struct mm_struct *prev,
> *
> * The "cpu" argument must be the current local CPU number.
> */
> -static inline void flush_icache_deferred(struct mm_struct *mm, unsigned int cpu)
> +static inline void flush_icache_deferred(struct mm_struct *mm, unsigned int cpu,
> + struct task_struct *task)
> {
> #ifdef CONFIG_SMP
> cpumask_t *mask = &mm->context.icache_stale_mask;
>
> - if (cpumask_test_cpu(cpu, mask)) {
> + if (cpumask_test_cpu(cpu, mask) || mm->context.force_icache_flush ||
> + (task && task->thread.force_icache_flush)) {
> cpumask_clear_cpu(cpu, mask);
> /*
> * Ensure the remote hart's writes are visible to this hart.
> @@ -332,5 +334,5 @@ void switch_mm(struct mm_struct *prev, struct mm_struct *next,
>
> set_mm(prev, next, cpu);
>
> - flush_icache_deferred(next, cpu);
> + flush_icache_deferred(next, cpu, task);
> }
> diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h
> index 370ed14b1ae0..472801ea78cc 100644
> --- a/include/uapi/linux/prctl.h
> +++ b/include/uapi/linux/prctl.h
> @@ -306,4 +306,7 @@ struct prctl_mm_map {
> # define PR_RISCV_V_VSTATE_CTRL_NEXT_MASK 0xc
> # define PR_RISCV_V_VSTATE_CTRL_MASK 0x1f
>
> +#define PR_RISCV_SET_ICACHE_FLUSH_CTX 71
> +# define PR_RISCV_CTX_SW_FENCEI 0
> +
> #endif /* _LINUX_PRCTL_H */
> diff --git a/kernel/sys.c b/kernel/sys.c
> index 420d9cb9cc8e..e806a8a67c36 100644
> --- a/kernel/sys.c
> +++ b/kernel/sys.c
> @@ -146,6 +146,9 @@
> #ifndef RISCV_V_GET_CONTROL
> # define RISCV_V_GET_CONTROL() (-EINVAL)
> #endif
> +#ifndef RISCV_SET_ICACHE_FLUSH_CTX
> +# define RISCV_SET_ICACHE_FLUSH_CTX(a, b) (-EINVAL)
> +#endif
>
> /*
> * this is where the system-wide overflow UID and GID are defined, for
> @@ -2739,6 +2742,9 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
> case PR_RISCV_V_GET_CONTROL:
> error = RISCV_V_GET_CONTROL();
> break;
> + case PR_RISCV_SET_ICACHE_FLUSH_CTX:
> + error = RISCV_SET_ICACHE_FLUSH_CTX(arg2, arg3);
> + break;
> default:
> error = -EINVAL;
> break;
>
> --
> 2.43.0
>
>
> _______________________________________________
> linux-riscv mailing list
> [email protected]
> http://lists.infradead.org/mailman/listinfo/linux-riscv
--
Regards,
Atish
On Fri, Jan 05, 2024 at 06:55:31PM -0800, Atish Patra wrote:
> On Wed, Dec 13, 2023 at 2:48 PM Charlie Jenkins <[email protected]> wrote:
> >
> > Support new prctl with key PR_RISCV_SET_ICACHE_FLUSH_CTX to enable
> > optimization of cross modifying code. This prctl enables userspace code
> > to use icache flushing instructions such as fence.i with the guarantee
> > that the icache will continue to be clean after thread migration.
> >
>
> Do we need a CLEAR option as well ? I am wondering if there is a use case where
> an application wants the prctl for only certain duration but may
> invoke fence.i multiple times (syscall will still not be a good option
> for them.)
Having a clear option does seem like a good idea. I will send out a new
patch with a context option that can be similar to
PR_RISCV_CTX_SW_FENCEI_CLEAR.
- Charlie
>
> > Signed-off-by: Charlie Jenkins <[email protected]>
> > ---
> > arch/riscv/include/asm/mmu.h | 2 ++
> > arch/riscv/include/asm/processor.h | 6 ++++++
> > arch/riscv/mm/cacheflush.c | 37 +++++++++++++++++++++++++++++++++++++
> > arch/riscv/mm/context.c | 8 +++++---
> > include/uapi/linux/prctl.h | 3 +++
> > kernel/sys.c | 6 ++++++
> > 6 files changed, 59 insertions(+), 3 deletions(-)
> >
> > diff --git a/arch/riscv/include/asm/mmu.h b/arch/riscv/include/asm/mmu.h
> > index 355504b37f8e..60be458e94da 100644
> > --- a/arch/riscv/include/asm/mmu.h
> > +++ b/arch/riscv/include/asm/mmu.h
> > @@ -19,6 +19,8 @@ typedef struct {
> > #ifdef CONFIG_SMP
> > /* A local icache flush is needed before user execution can resume. */
> > cpumask_t icache_stale_mask;
> > + /* Force local icache flush on all migrations. */
> > + bool force_icache_flush;
> > #endif
> > #ifdef CONFIG_BINFMT_ELF_FDPIC
> > unsigned long exec_fdpic_loadmap;
> > diff --git a/arch/riscv/include/asm/processor.h b/arch/riscv/include/asm/processor.h
> > index f19f861cda54..7eda6c75e0f2 100644
> > --- a/arch/riscv/include/asm/processor.h
> > +++ b/arch/riscv/include/asm/processor.h
> > @@ -84,6 +84,9 @@ struct thread_struct {
> > unsigned long vstate_ctrl;
> > struct __riscv_v_ext_state vstate;
> > unsigned long align_ctl;
> > +#ifdef CONFIG_SMP
> > + bool force_icache_flush;
> > +#endif
> > };
> >
> > /* Whitelist the fstate from the task_struct for hardened usercopy */
> > @@ -145,6 +148,9 @@ extern int set_unalign_ctl(struct task_struct *tsk, unsigned int val);
> > #define GET_UNALIGN_CTL(tsk, addr) get_unalign_ctl((tsk), (addr))
> > #define SET_UNALIGN_CTL(tsk, val) set_unalign_ctl((tsk), (val))
> >
> > +#define RISCV_SET_ICACHE_FLUSH_CTX(arg1, arg2) riscv_set_icache_flush_ctx(arg1, arg2)
> > +extern int riscv_set_icache_flush_ctx(unsigned long ctx, unsigned long per_thread);
> > +
> > #endif /* __ASSEMBLY__ */
> >
> > #endif /* _ASM_RISCV_PROCESSOR_H */
> > diff --git a/arch/riscv/mm/cacheflush.c b/arch/riscv/mm/cacheflush.c
> > index 55a34f2020a8..3b2bf8256a10 100644
> > --- a/arch/riscv/mm/cacheflush.c
> > +++ b/arch/riscv/mm/cacheflush.c
> > @@ -5,6 +5,7 @@
> >
> > #include <linux/acpi.h>
> > #include <linux/of.h>
> > +#include <linux/prctl.h>
> > #include <asm/acpi.h>
> > #include <asm/cacheflush.h>
> >
> > @@ -152,3 +153,39 @@ void __init riscv_init_cbo_blocksizes(void)
> > if (cboz_block_size)
> > riscv_cboz_block_size = cboz_block_size;
> > }
> > +
> > +/**
> > + * riscv_set_icache_flush_ctx() - Enable userspace to emit icache flushing instructions.
> > + * @ctx: Sets the type of context
> > + * - PR_RISCV_CTX_SW_FENCEI: Allow fence.i in userspace. Another fence.i will
> > + * emitted on thread/process migration.
> > + * @per_thread: When set to 0, will use the default behavior of setting the
> > + * icache flush context per process. When set to 1, will use a per thread
> > + * context.
> > + *
> > + * When in per-process context, there may be multiple threads using the same mm.
> > + * Therefore, the icache can never be assumed clean when. Multiple threads in
> > + * the process may modify instructions in the mm concurrently.
> > + *
> > + * In per-thread context, it can be assumed that all modifications to
> > + * instructions in memory will be performed by this thread. When the thread is
> > + * migrated the icache will be flushed.
> > + *
> > + */
> > +int riscv_set_icache_flush_ctx(unsigned long ctx, unsigned long per_thread)
> > +{
> > +#ifdef CONFIG_SMP
> > + switch (ctx) {
> > + case PR_RISCV_CTX_SW_FENCEI:
> > + if (per_thread)
> > + current->thread.force_icache_flush = true;
> > + else
> > + current->mm->context.force_icache_flush = true;
> > + break;
> > +
> > + default:
> > + break;
> > + }
> > +#endif
> > + return 0;
> > +}
> > diff --git a/arch/riscv/mm/context.c b/arch/riscv/mm/context.c
> > index 217fd4de6134..0146c61be0ab 100644
> > --- a/arch/riscv/mm/context.c
> > +++ b/arch/riscv/mm/context.c
> > @@ -297,12 +297,14 @@ static inline void set_mm(struct mm_struct *prev,
> > *
> > * The "cpu" argument must be the current local CPU number.
> > */
> > -static inline void flush_icache_deferred(struct mm_struct *mm, unsigned int cpu)
> > +static inline void flush_icache_deferred(struct mm_struct *mm, unsigned int cpu,
> > + struct task_struct *task)
> > {
> > #ifdef CONFIG_SMP
> > cpumask_t *mask = &mm->context.icache_stale_mask;
> >
> > - if (cpumask_test_cpu(cpu, mask)) {
> > + if (cpumask_test_cpu(cpu, mask) || mm->context.force_icache_flush ||
> > + (task && task->thread.force_icache_flush)) {
> > cpumask_clear_cpu(cpu, mask);
> > /*
> > * Ensure the remote hart's writes are visible to this hart.
> > @@ -332,5 +334,5 @@ void switch_mm(struct mm_struct *prev, struct mm_struct *next,
> >
> > set_mm(prev, next, cpu);
> >
> > - flush_icache_deferred(next, cpu);
> > + flush_icache_deferred(next, cpu, task);
> > }
> > diff --git a/include/uapi/linux/prctl.h b/include/uapi/linux/prctl.h
> > index 370ed14b1ae0..472801ea78cc 100644
> > --- a/include/uapi/linux/prctl.h
> > +++ b/include/uapi/linux/prctl.h
> > @@ -306,4 +306,7 @@ struct prctl_mm_map {
> > # define PR_RISCV_V_VSTATE_CTRL_NEXT_MASK 0xc
> > # define PR_RISCV_V_VSTATE_CTRL_MASK 0x1f
> >
> > +#define PR_RISCV_SET_ICACHE_FLUSH_CTX 71
> > +# define PR_RISCV_CTX_SW_FENCEI 0
> > +
> > #endif /* _LINUX_PRCTL_H */
> > diff --git a/kernel/sys.c b/kernel/sys.c
> > index 420d9cb9cc8e..e806a8a67c36 100644
> > --- a/kernel/sys.c
> > +++ b/kernel/sys.c
> > @@ -146,6 +146,9 @@
> > #ifndef RISCV_V_GET_CONTROL
> > # define RISCV_V_GET_CONTROL() (-EINVAL)
> > #endif
> > +#ifndef RISCV_SET_ICACHE_FLUSH_CTX
> > +# define RISCV_SET_ICACHE_FLUSH_CTX(a, b) (-EINVAL)
> > +#endif
> >
> > /*
> > * this is where the system-wide overflow UID and GID are defined, for
> > @@ -2739,6 +2742,9 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
> > case PR_RISCV_V_GET_CONTROL:
> > error = RISCV_V_GET_CONTROL();
> > break;
> > + case PR_RISCV_SET_ICACHE_FLUSH_CTX:
> > + error = RISCV_SET_ICACHE_FLUSH_CTX(arg2, arg3);
> > + break;
> > default:
> > error = -EINVAL;
> > break;
> >
> > --
> > 2.43.0
> >
> >
> > _______________________________________________
> > linux-riscv mailing list
> > [email protected]
> > http://lists.infradead.org/mailman/listinfo/linux-riscv
>
>
>
> --
> Regards,
> Atish
On Fri, Jan 05, 2024 at 06:52:12PM -0800, Atish Patra wrote:
> On Wed, Dec 13, 2023 at 2:48 PM Charlie Jenkins <[email protected]> wrote:
> >
> > Provide documentation that explains how to properly do CMODX in riscv.
> >
> > Signed-off-by: Charlie Jenkins <[email protected]>
> > ---
> > Documentation/arch/riscv/cmodx.rst | 98 ++++++++++++++++++++++++++++++++++++++
> > Documentation/arch/riscv/index.rst | 1 +
> > 2 files changed, 99 insertions(+)
> >
> > diff --git a/Documentation/arch/riscv/cmodx.rst b/Documentation/arch/riscv/cmodx.rst
> > new file mode 100644
> > index 000000000000..20f327d85116
> > --- /dev/null
> > +++ b/Documentation/arch/riscv/cmodx.rst
> > @@ -0,0 +1,98 @@
> > +.. SPDX-License-Identifier: GPL-2.0
> > +
> > +==============================================================================
> > +Concurrent Modification and Execution of Instructions (CMODX) for RISC-V Linux
> > +==============================================================================
> > +
> > +CMODX is a programming technique where a program executes instructions that were
> > +modified by the program itself. Instruction storage and the instruction cache
> > +(icache) is not guaranteed to be synchronized on RISC-V hardware. Therefore, the
> > +program must enforce its own synchonization with the unprivileged fence.i/
>
> /s/synchonization/synchronization
>
> > +instruction.
> > +
> > +However, the default Linux ABI prohibits the use of fence.i in userspace
> > +applications. At any point the scheduler may migrate a task onto a new hart. If
> > +migration occurs after the userspace synchronized the icache and instruction
> > +storage with fence.i, the icache will no longer be clean. This is due to the
> > +behavior of fence.i only affecting the hart that it is called on. Thus, the hart
> > +that the task has been migrated to, may not have synchronized instruction
> > +storage and icache.
> > +
> > +There are two ways to solve this problem: use the riscv_flush_icache() syscall,
> > +or use the ``PR_RISCV_SET_ICACHE_FLUSH_CTX`` prctl(). The syscall should be used
> > +when the application very rarely needs to flush the icache. If the icache will
>
> The syscall is a one time operation while prctl is sticky.
> It would be great if we can add a little more context why the syscall
> behaves this way compared to prctl.
I can highlight the point that the prctl is sticky and the syscall is
not.
As for "why", they simply fill different roles. It is useful to have
both a sticky and a non-sticky option. I chose the sticky operation to
be a prctl rather than a generic syscall because the semantics of prctl
is that they operate on process or thread scoped behavior which is what
was needed.
- Charlie
>
> > +need to be flushed many times in the lifetime of the application, the prctl
> > +should be used.
> > +
> > +The prctl informs the kernel that it must emit synchronizing instructions upon
> > +task migration. The program itself must emit synchonizing instructions when
>
> /s/synchonizing/synchronizing
>
> > +necessary as well.
> > +
> > +1. prctl() Interface
> > +---------------------
> > +
> > +Before the program emits their first icache flushing instruction, the program
> > +must call this prctl().
> > +
> > +* prctl(PR_RISCV_SET_ICACHE_FLUSH_CTX, unsigned long ctx, unsigned long per_thread)
> > +
> > + Sets the icache flushing context. If per_thread is 0, context will be
> > + applied per process, otherwise if per_thread is 1 context will be
> > + per-thread. Any other number will have undefined behavior.
> > +
> > + * :c:macro:`PR_RISCV_CTX_SW_FENCEI`: Allow fence.i to be called in
> > + userspace.
> > +
> > +Example usage:
> > +
> > +The following files are meant to be compiled and linked with each other. The
> > +modify_instruction() function replaces an add with 0 with an add with one,
> > +causing the instruction sequence in get_value() to change from returning a zero
> > +to returning a one.
> > +
> > +cmodx.c::
> > +
> > + #include <stdio.h>
> > + #include <sys/prctl.h>
> > +
> > + extern int get_value();
> > + extern void modify_instruction();
> > +
> > + int main()
> > + {
> > + int value = get_value();
> > + printf("Value before cmodx: %d\n", value);
> > +
> > + // Call prctl before first fence.i is called inside modify_instruction
> > + prctl(PR_RISCV_SET_ICACHE_FLUSH_CTX, PR_RISCV_CTX_SW_FENCEI, 0);
> > + modify_instruction();
> > +
> > + value = get_value();
> > + printf("Value after cmodx: %d\n", value);
> > + return 0;
> > + }
> > +
> > +cmodx.S::
> > +
> > + .option norvc
> > +
> > + .text
> > + .global modify_instruction
> > + modify_instruction:
> > + lw a0, new_insn
> > + lui a5,%hi(old_insn)
> > + sw a0,%lo(old_insn)(a5)
> > + fence.i
> > + ret
> > +
> > + .section modifiable, "awx"
> > + .global get_value
> > + get_value:
> > + li a0, 0
> > + old_insn:
> > + addi a0, a0, 0
> > + ret
> > +
> > + .data
> > + new_insn:
> > + addi a0, a0, 1
> > diff --git a/Documentation/arch/riscv/index.rst b/Documentation/arch/riscv/index.rst
> > index 4dab0cb4b900..eecf347ce849 100644
> > --- a/Documentation/arch/riscv/index.rst
> > +++ b/Documentation/arch/riscv/index.rst
> > @@ -13,6 +13,7 @@ RISC-V architecture
> > patch-acceptance
> > uabi
> > vector
> > + cmodx
> >
> > features
> >
> >
> > --
> > 2.43.0
> >
> >
> > _______________________________________________
> > linux-riscv mailing list
> > [email protected]
> > http://lists.infradead.org/mailman/listinfo/linux-riscv
>
>
>
> --
> Regards,
> Atish
On Fri, Jan 5, 2024 at 7:52 PM Charlie Jenkins <[email protected]> wrote:
>
> On Fri, Jan 05, 2024 at 06:52:12PM -0800, Atish Patra wrote:
> > On Wed, Dec 13, 2023 at 2:48 PM Charlie Jenkins <[email protected]> wrote:
> > >
> > > Provide documentation that explains how to properly do CMODX in riscv.
> > >
> > > Signed-off-by: Charlie Jenkins <[email protected]>
> > > ---
> > > Documentation/arch/riscv/cmodx.rst | 98 ++++++++++++++++++++++++++++++++++++++
> > > Documentation/arch/riscv/index.rst | 1 +
> > > 2 files changed, 99 insertions(+)
> > >
> > > diff --git a/Documentation/arch/riscv/cmodx.rst b/Documentation/arch/riscv/cmodx.rst
> > > new file mode 100644
> > > index 000000000000..20f327d85116
> > > --- /dev/null
> > > +++ b/Documentation/arch/riscv/cmodx.rst
> > > @@ -0,0 +1,98 @@
> > > +.. SPDX-License-Identifier: GPL-2.0
> > > +
> > > +==============================================================================
> > > +Concurrent Modification and Execution of Instructions (CMODX) for RISC-V Linux
> > > +==============================================================================
> > > +
> > > +CMODX is a programming technique where a program executes instructions that were
> > > +modified by the program itself. Instruction storage and the instruction cache
> > > +(icache) is not guaranteed to be synchronized on RISC-V hardware. Therefore, the
> > > +program must enforce its own synchonization with the unprivileged fence.i/
> >
> > /s/synchonization/synchronization
> >
> > > +instruction.
> > > +
> > > +However, the default Linux ABI prohibits the use of fence.i in userspace
> > > +applications. At any point the scheduler may migrate a task onto a new hart. If
> > > +migration occurs after the userspace synchronized the icache and instruction
> > > +storage with fence.i, the icache will no longer be clean. This is due to the
> > > +behavior of fence.i only affecting the hart that it is called on. Thus, the hart
> > > +that the task has been migrated to, may not have synchronized instruction
> > > +storage and icache.
> > > +
> > > +There are two ways to solve this problem: use the riscv_flush_icache() syscall,
> > > +or use the ``PR_RISCV_SET_ICACHE_FLUSH_CTX`` prctl(). The syscall should be used
> > > +when the application very rarely needs to flush the icache. If the icache will
> >
> > The syscall is a one time operation while prctl is sticky.
> > It would be great if we can add a little more context why the syscall
> > behaves this way compared to prctl.
>
> I can highlight the point that the prctl is sticky and the syscall is
> not.
>
> As for "why", they simply fill different roles. It is useful to have
> both a sticky and a non-sticky option. I chose the sticky operation to
> be a prctl rather than a generic syscall because the semantics of prctl
> is that they operate on process or thread scoped behavior which is what
> was needed.
>
Yeah. That's what I was suggesting. It is enough for readers to understand
the difference between the two approaches without reading the kernel
source code.
> - Charlie
>
> >
> > > +need to be flushed many times in the lifetime of the application, the prctl
> > > +should be used.
> > > +
> > > +The prctl informs the kernel that it must emit synchronizing instructions upon
> > > +task migration. The program itself must emit synchonizing instructions when
> >
> > /s/synchonizing/synchronizing
> >
> > > +necessary as well.
> > > +
> > > +1. prctl() Interface
> > > +---------------------
> > > +
> > > +Before the program emits their first icache flushing instruction, the program
> > > +must call this prctl().
> > > +
> > > +* prctl(PR_RISCV_SET_ICACHE_FLUSH_CTX, unsigned long ctx, unsigned long per_thread)
> > > +
> > > + Sets the icache flushing context. If per_thread is 0, context will be
> > > + applied per process, otherwise if per_thread is 1 context will be
> > > + per-thread. Any other number will have undefined behavior.
> > > +
> > > + * :c:macro:`PR_RISCV_CTX_SW_FENCEI`: Allow fence.i to be called in
> > > + userspace.
> > > +
> > > +Example usage:
> > > +
> > > +The following files are meant to be compiled and linked with each other. The
> > > +modify_instruction() function replaces an add with 0 with an add with one,
> > > +causing the instruction sequence in get_value() to change from returning a zero
> > > +to returning a one.
> > > +
> > > +cmodx.c::
> > > +
> > > + #include <stdio.h>
> > > + #include <sys/prctl.h>
> > > +
> > > + extern int get_value();
> > > + extern void modify_instruction();
> > > +
> > > + int main()
> > > + {
> > > + int value = get_value();
> > > + printf("Value before cmodx: %d\n", value);
> > > +
> > > + // Call prctl before first fence.i is called inside modify_instruction
> > > + prctl(PR_RISCV_SET_ICACHE_FLUSH_CTX, PR_RISCV_CTX_SW_FENCEI, 0);
> > > + modify_instruction();
> > > +
> > > + value = get_value();
> > > + printf("Value after cmodx: %d\n", value);
> > > + return 0;
> > > + }
> > > +
> > > +cmodx.S::
> > > +
> > > + .option norvc
> > > +
> > > + .text
> > > + .global modify_instruction
> > > + modify_instruction:
> > > + lw a0, new_insn
> > > + lui a5,%hi(old_insn)
> > > + sw a0,%lo(old_insn)(a5)
> > > + fence.i
> > > + ret
> > > +
> > > + .section modifiable, "awx"
> > > + .global get_value
> > > + get_value:
> > > + li a0, 0
> > > + old_insn:
> > > + addi a0, a0, 0
> > > + ret
> > > +
> > > + .data
> > > + new_insn:
> > > + addi a0, a0, 1
> > > diff --git a/Documentation/arch/riscv/index.rst b/Documentation/arch/riscv/index.rst
> > > index 4dab0cb4b900..eecf347ce849 100644
> > > --- a/Documentation/arch/riscv/index.rst
> > > +++ b/Documentation/arch/riscv/index.rst
> > > @@ -13,6 +13,7 @@ RISC-V architecture
> > > patch-acceptance
> > > uabi
> > > vector
> > > + cmodx
> > >
> > > features
> > >
> > >
> > > --
> > > 2.43.0
> > >
> > >
> > > _______________________________________________
> > > linux-riscv mailing list
> > > [email protected]
> > > http://lists.infradead.org/mailman/listinfo/linux-riscv
> >
> >
> >
> > --
> > Regards,
> > Atish
--
Regards,
Atish