[v1] riscv/kprobes: allocate detour buffer from module area

[PATCH] riscv/kprobes: allocate detour buffer from module area

Posted by Liao Chang 3 years, 9 months ago

To address the limitation of PC-relative branch instruction
on riscv architecture, detour buffer slot is allocated from
a area, the distance of which from kernel should be less than 4GB.

For the time being, Modules region always live before the kernel.
But Vmalloc region reside far away from kernel, the distance is
half of the kernel address space.

Signed-off-by: Liao Chang <liaochang1@huawei.com>
---
 arch/riscv/kernel/probes/kprobes.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/arch/riscv/kernel/probes/kprobes.c b/arch/riscv/kernel/probes/kprobes.c
index e6e950b7cf32..bc027a663b17 100644
--- a/arch/riscv/kernel/probes/kprobes.c
+++ b/arch/riscv/kernel/probes/kprobes.c
@@ -6,12 +6,14 @@
 #include <linux/extable.h>
 #include <linux/slab.h>
 #include <linux/stop_machine.h>
+#include <linux/moduleloader.h>
 #include <asm/ptrace.h>
 #include <linux/uaccess.h>
 #include <asm/sections.h>
 #include <asm/cacheflush.h>
 #include <asm/bug.h>
 #include <asm/patch.h>
+#include <asm/set_memory.h>
 
 #include "decode-insn.h"
 
@@ -86,10 +88,28 @@ int __kprobes arch_prepare_kprobe(struct kprobe *p)
 #ifdef CONFIG_MMU
 void *alloc_insn_page(void)
 {
+#if defined(CONFIG_MODULES) && defined(CONFIG_64BIT)
+	void *page;
+
+	page = module_alloc(PAGE_SIZE);
+	if (!page)
+		return NULL;
+
+	set_vm_flush_reset_perms(page);
+	/*
+	 * First make the page read-only, and only then make it executable to
+	 * prevent it from being W+X in between.
+	 */
+	set_memory_ro((unsigned long)page, 1);
+	set_memory_x((unsigned long)page, 1);
+
+	return page;
+#else
 	return  __vmalloc_node_range(PAGE_SIZE, 1, VMALLOC_START, VMALLOC_END,
 				     GFP_KERNEL, PAGE_KERNEL_READ_EXEC,
 				     VM_FLUSH_RESET_PERMS, NUMA_NO_NODE,
 				     __builtin_return_address(0));
+#endif
 }
 #endif
 
-- 
2.17.1

Re: [PATCH] riscv/kprobes: allocate detour buffer from module area

Posted by Palmer Dabbelt 3 years, 8 months ago

On Tue, 28 Jun 2022 18:13:17 PDT (-0700), liaochang1@huawei.com wrote:
> To address the limitation of PC-relative branch instruction
> on riscv architecture, detour buffer slot is allocated from
> a area, the distance of which from kernel should be less than 4GB.
>
> For the time being, Modules region always live before the kernel.
> But Vmalloc region reside far away from kernel, the distance is
> half of the kernel address space.
>
> Signed-off-by: Liao Chang <liaochang1@huawei.com>
> ---
>  arch/riscv/kernel/probes/kprobes.c | 20 ++++++++++++++++++++
>  1 file changed, 20 insertions(+)
>
> diff --git a/arch/riscv/kernel/probes/kprobes.c b/arch/riscv/kernel/probes/kprobes.c
> index e6e950b7cf32..bc027a663b17 100644
> --- a/arch/riscv/kernel/probes/kprobes.c
> +++ b/arch/riscv/kernel/probes/kprobes.c
> @@ -6,12 +6,14 @@
>  #include <linux/extable.h>
>  #include <linux/slab.h>
>  #include <linux/stop_machine.h>
> +#include <linux/moduleloader.h>
>  #include <asm/ptrace.h>
>  #include <linux/uaccess.h>
>  #include <asm/sections.h>
>  #include <asm/cacheflush.h>
>  #include <asm/bug.h>
>  #include <asm/patch.h>
> +#include <asm/set_memory.h>
>
>  #include "decode-insn.h"
>
> @@ -86,10 +88,28 @@ int __kprobes arch_prepare_kprobe(struct kprobe *p)
>  #ifdef CONFIG_MMU
>  void *alloc_insn_page(void)
>  {
> +#if defined(CONFIG_MODULES) && defined(CONFIG_64BIT)
> +	void *page;
> +
> +	page = module_alloc(PAGE_SIZE);
> +	if (!page)
> +		return NULL;
> +
> +	set_vm_flush_reset_perms(page);
> +	/*
> +	 * First make the page read-only, and only then make it executable to
> +	 * prevent it from being W+X in between.
> +	 */
> +	set_memory_ro((unsigned long)page, 1);
> +	set_memory_x((unsigned long)page, 1);
> +
> +	return page;
> +#else
>  	return  __vmalloc_node_range(PAGE_SIZE, 1, VMALLOC_START, VMALLOC_END,
>  				     GFP_KERNEL, PAGE_KERNEL_READ_EXEC,
>  				     VM_FLUSH_RESET_PERMS, NUMA_NO_NODE,
>  				     __builtin_return_address(0));
> +#endif
>  }
>  #endif

This probably shouldn't depend on CONFIG_MODULES, it's just as broken to 
assume we can jump to the vmalloc region regardless of whether modules 
are enabled or disabled.  We should be able to just allocate from the 
jump-safe region either way.

Re: [PATCH] riscv/kprobes: allocate detour buffer from module area

Posted by liaochang (A) 3 years, 7 months ago


在 2022/8/12 1:09, Palmer Dabbelt 写道:
> On Tue, 28 Jun 2022 18:13:17 PDT (-0700), liaochang1@huawei.com wrote:
>> To address the limitation of PC-relative branch instruction
>> on riscv architecture, detour buffer slot is allocated from
>> a area, the distance of which from kernel should be less than 4GB.
>>
>> For the time being, Modules region always live before the kernel.
>> But Vmalloc region reside far away from kernel, the distance is
>> half of the kernel address space.
>>
>> Signed-off-by: Liao Chang <liaochang1@huawei.com>
>> ---
>>  arch/riscv/kernel/probes/kprobes.c | 20 ++++++++++++++++++++
>>  1 file changed, 20 insertions(+)
>>
>> diff --git a/arch/riscv/kernel/probes/kprobes.c b/arch/riscv/kernel/probes/kprobes.c
>> index e6e950b7cf32..bc027a663b17 100644
>> --- a/arch/riscv/kernel/probes/kprobes.c
>> +++ b/arch/riscv/kernel/probes/kprobes.c
>> @@ -6,12 +6,14 @@
>>  #include <linux/extable.h>
>>  #include <linux/slab.h>
>>  #include <linux/stop_machine.h>
>> +#include <linux/moduleloader.h>
>>  #include <asm/ptrace.h>
>>  #include <linux/uaccess.h>
>>  #include <asm/sections.h>
>>  #include <asm/cacheflush.h>
>>  #include <asm/bug.h>
>>  #include <asm/patch.h>
>> +#include <asm/set_memory.h>
>>
>>  #include "decode-insn.h"
>>
>> @@ -86,10 +88,28 @@ int __kprobes arch_prepare_kprobe(struct kprobe *p)
>>  #ifdef CONFIG_MMU
>>  void *alloc_insn_page(void)
>>  {
>> +#if defined(CONFIG_MODULES) && defined(CONFIG_64BIT)
>> +    void *page;
>> +
>> +    page = module_alloc(PAGE_SIZE);
>> +    if (!page)
>> +        return NULL;
>> +
>> +    set_vm_flush_reset_perms(page);
>> +    /*
>> +     * First make the page read-only, and only then make it executable to
>> +     * prevent it from being W+X in between.
>> +     */
>> +    set_memory_ro((unsigned long)page, 1);
>> +    set_memory_x((unsigned long)page, 1);
>> +
>> +    return page;
>> +#else
>>      return  __vmalloc_node_range(PAGE_SIZE, 1, VMALLOC_START, VMALLOC_END,
>>                       GFP_KERNEL, PAGE_KERNEL_READ_EXEC,
>>                       VM_FLUSH_RESET_PERMS, NUMA_NO_NODE,
>>                       __builtin_return_address(0));
>> +#endif
>>  }
>>  #endif
> 
> This probably shouldn't depend on CONFIG_MODULES, it's just as broken to assume we can jump to the vmalloc region regardless of whether modules are enabled or disabled.  We should be able to just allocate from the jump-safe region either way.
> .

Yes, there is no need to invoke module_alloc to allocate buffer from jump-safe region,
just use __vmalloc_node_range directly, thanks for reminding me.

-- 
BR,
Liao, Chang