diff mbox series

[v4] powerpc/uprobes: Validation for prefixed instruction

Message ID 20210305115433.140769-1-ravi.bangoria@linux.ibm.com (mailing list archive)
State Changes Requested
Headers show
Series [v4] powerpc/uprobes: Validation for prefixed instruction | expand
Related show

Checks

Context Check Description
snowpatch_ozlabs/apply_patch success Successfully applied on branch powerpc/merge (91966823812efbd175f904599e5cf2a854b39809)
snowpatch_ozlabs/build-ppc64le success Build succeeded
snowpatch_ozlabs/build-ppc64be success Build succeeded
snowpatch_ozlabs/build-ppc64e success Build succeeded
snowpatch_ozlabs/build-pmac32 success Build succeeded
snowpatch_ozlabs/checkpatch success total: 0 errors, 0 warnings, 0 checks, 13 lines checked
snowpatch_ozlabs/needsstable success Patch has no Fixes tags

Commit Message

Ravi Bangoria March 5, 2021, 11:54 a.m. UTC 
As per ISA 3.1, prefixed instruction should not cross 64-byte
boundary. So don't allow Uprobe on such prefixed instruction.

There are two ways probed instruction is changed in mapped pages.
First, when Uprobe is activated, it searches for all the relevant
pages and replace instruction in them. In this case, if that probe
is on the 64-byte unaligned prefixed instruction, error out
directly. Second, when Uprobe is already active and user maps a
relevant page via mmap(), instruction is replaced via mmap() code
path. But because Uprobe is invalid, entire mmap() operation can
not be stopped. In this case just print an error and continue.

Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
---
v3: https://lore.kernel.org/r/20210304050529.59391-1-ravi.bangoria@linux.ibm.com
v3->v4:
  - CONFIG_PPC64 check was not required, remove it.
  - Use SZ_ macros instead of hardcoded numbers.

 arch/powerpc/kernel/uprobes.c | 7 +++++++
 1 file changed, 7 insertions(+)

Comments

Sandipan Das March 8, 2021, 4:33 a.m. UTC | #1 
On 05/03/21 5:24 pm, Ravi Bangoria wrote:
> As per ISA 3.1, prefixed instruction should not cross 64-byte
> boundary. So don't allow Uprobe on such prefixed instruction.
> 
> There are two ways probed instruction is changed in mapped pages.
> First, when Uprobe is activated, it searches for all the relevant
> pages and replace instruction in them. In this case, if that probe
> is on the 64-byte unaligned prefixed instruction, error out
> directly. Second, when Uprobe is already active and user maps a
> relevant page via mmap(), instruction is replaced via mmap() code
> path. But because Uprobe is invalid, entire mmap() operation can
> not be stopped. In this case just print an error and continue.
> 
> Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
> Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
> ---
> v3: https://lore.kernel.org/r/20210304050529.59391-1-ravi.bangoria@linux.ibm.com
> v3->v4:
>   - CONFIG_PPC64 check was not required, remove it.
>   - Use SZ_ macros instead of hardcoded numbers.
> 

Acked-by: Sandipan Das <sandipan@linux.ibm.com>
Michael Ellerman March 9, 2021, 9:54 a.m. UTC | #2 
Ravi Bangoria <ravi.bangoria@linux.ibm.com> writes:
> As per ISA 3.1, prefixed instruction should not cross 64-byte
> boundary. So don't allow Uprobe on such prefixed instruction.
>
> There are two ways probed instruction is changed in mapped pages.
> First, when Uprobe is activated, it searches for all the relevant
> pages and replace instruction in them. In this case, if that probe
> is on the 64-byte unaligned prefixed instruction, error out
> directly. Second, when Uprobe is already active and user maps a
> relevant page via mmap(), instruction is replaced via mmap() code
> path. But because Uprobe is invalid, entire mmap() operation can
> not be stopped. In this case just print an error and continue.
>
> Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
> Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>

Do we have a Fixes: tag for this?

> ---
> v3: https://lore.kernel.org/r/20210304050529.59391-1-ravi.bangoria@linux.ibm.com
> v3->v4:
>   - CONFIG_PPC64 check was not required, remove it.
>   - Use SZ_ macros instead of hardcoded numbers.
>
>  arch/powerpc/kernel/uprobes.c | 7 +++++++
>  1 file changed, 7 insertions(+)
>
> diff --git a/arch/powerpc/kernel/uprobes.c b/arch/powerpc/kernel/uprobes.c
> index e8a63713e655..4cbfff6e94a3 100644
> --- a/arch/powerpc/kernel/uprobes.c
> +++ b/arch/powerpc/kernel/uprobes.c
> @@ -41,6 +41,13 @@ int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe,
>  	if (addr & 0x03)
>  		return -EINVAL;
>  
> +	if (cpu_has_feature(CPU_FTR_ARCH_31) &&
> +	    ppc_inst_prefixed(auprobe->insn) &&
> +	    (addr & (SZ_64 - 4)) == SZ_64 - 4) {
> +		pr_info_ratelimited("Cannot register a uprobe on 64 byte unaligned prefixed instruction\n");
> +		return -EINVAL;

I realise we already did the 0x03 check above, but I still think this
would be clearer simply as:

	    (addr & 0x3f == 60)

cheers
Naveen N. Rao March 9, 2021, 11:21 a.m. UTC | #3 
On 2021/03/09 08:54PM, Michael Ellerman wrote:
> Ravi Bangoria <ravi.bangoria@linux.ibm.com> writes:
> > As per ISA 3.1, prefixed instruction should not cross 64-byte
> > boundary. So don't allow Uprobe on such prefixed instruction.
> >
> > There are two ways probed instruction is changed in mapped pages.
> > First, when Uprobe is activated, it searches for all the relevant
> > pages and replace instruction in them. In this case, if that probe
> > is on the 64-byte unaligned prefixed instruction, error out
> > directly. Second, when Uprobe is already active and user maps a
> > relevant page via mmap(), instruction is replaced via mmap() code
> > path. But because Uprobe is invalid, entire mmap() operation can
> > not be stopped. In this case just print an error and continue.
> >
> > Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
> > Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
> 
> Do we have a Fixes: tag for this?

Since this is an additional check we are adding, I don't think we should 
add a Fixes: tag. Nothing is broken per-se -- we're just adding more 
checks to catch simple mistakes. Also, like Oleg pointed out, there are 
still many other ways for users to shoot themselves in the foot with 
uprobes and prefixed instructions, if they so desire.

However, if you still think we should add a Fixes: tag, we can perhaps 
use the below commit since I didn't see any specific commit adding 
support for prefixed instructions for uprobes:

Fixes: 650b55b707fdfa ("powerpc: Add prefixed instructions to 
instruction data type")

> 
> > ---
> > v3: https://lore.kernel.org/r/20210304050529.59391-1-ravi.bangoria@linux.ibm.com
> > v3->v4:
> >   - CONFIG_PPC64 check was not required, remove it.
> >   - Use SZ_ macros instead of hardcoded numbers.
> >
> >  arch/powerpc/kernel/uprobes.c | 7 +++++++
> >  1 file changed, 7 insertions(+)
> >
> > diff --git a/arch/powerpc/kernel/uprobes.c b/arch/powerpc/kernel/uprobes.c
> > index e8a63713e655..4cbfff6e94a3 100644
> > --- a/arch/powerpc/kernel/uprobes.c
> > +++ b/arch/powerpc/kernel/uprobes.c
> > @@ -41,6 +41,13 @@ int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe,
> >  	if (addr & 0x03)
> >  		return -EINVAL;
> >  
> > +	if (cpu_has_feature(CPU_FTR_ARCH_31) &&
> > +	    ppc_inst_prefixed(auprobe->insn) &&
> > +	    (addr & (SZ_64 - 4)) == SZ_64 - 4) {
> > +		pr_info_ratelimited("Cannot register a uprobe on 64 byte unaligned prefixed instruction\n");
> > +		return -EINVAL;
> 
> I realise we already did the 0x03 check above, but I still think this
> would be clearer simply as:
> 
> 	    (addr & 0x3f == 60)

Indeed, I like the use of `60' there -- hex is overrated ;)

- Naveen
Ravi Bangoria March 9, 2021, 12:58 p.m. UTC | #4 
On 3/9/21 4:51 PM, Naveen N. Rao wrote:
> On 2021/03/09 08:54PM, Michael Ellerman wrote:
>> Ravi Bangoria <ravi.bangoria@linux.ibm.com> writes:
>>> As per ISA 3.1, prefixed instruction should not cross 64-byte
>>> boundary. So don't allow Uprobe on such prefixed instruction.
>>>
>>> There are two ways probed instruction is changed in mapped pages.
>>> First, when Uprobe is activated, it searches for all the relevant
>>> pages and replace instruction in them. In this case, if that probe
>>> is on the 64-byte unaligned prefixed instruction, error out
>>> directly. Second, when Uprobe is already active and user maps a
>>> relevant page via mmap(), instruction is replaced via mmap() code
>>> path. But because Uprobe is invalid, entire mmap() operation can
>>> not be stopped. In this case just print an error and continue.
>>>
>>> Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
>>> Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
>>
>> Do we have a Fixes: tag for this?
> 
> Since this is an additional check we are adding, I don't think we should
> add a Fixes: tag. Nothing is broken per-se -- we're just adding more
> checks to catch simple mistakes. Also, like Oleg pointed out, there are
> still many other ways for users to shoot themselves in the foot with
> uprobes and prefixed instructions, if they so desire.
> 
> However, if you still think we should add a Fixes: tag, we can perhaps
> use the below commit since I didn't see any specific commit adding
> support for prefixed instructions for uprobes:
> 
> Fixes: 650b55b707fdfa ("powerpc: Add prefixed instructions to
> instruction data type")

True. IMO, It doesn't really need any Fixes tag.

> 
>>
>>> ---
>>> v3: https://lore.kernel.org/r/20210304050529.59391-1-ravi.bangoria@linux.ibm.com
>>> v3->v4:
>>>    - CONFIG_PPC64 check was not required, remove it.
>>>    - Use SZ_ macros instead of hardcoded numbers.
>>>
>>>   arch/powerpc/kernel/uprobes.c | 7 +++++++
>>>   1 file changed, 7 insertions(+)
>>>
>>> diff --git a/arch/powerpc/kernel/uprobes.c b/arch/powerpc/kernel/uprobes.c
>>> index e8a63713e655..4cbfff6e94a3 100644
>>> --- a/arch/powerpc/kernel/uprobes.c
>>> +++ b/arch/powerpc/kernel/uprobes.c
>>> @@ -41,6 +41,13 @@ int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe,
>>>   	if (addr & 0x03)
>>>   		return -EINVAL;
>>>   
>>> +	if (cpu_has_feature(CPU_FTR_ARCH_31) &&
>>> +	    ppc_inst_prefixed(auprobe->insn) &&
>>> +	    (addr & (SZ_64 - 4)) == SZ_64 - 4) {
>>> +		pr_info_ratelimited("Cannot register a uprobe on 64 byte unaligned prefixed instruction\n");
>>> +		return -EINVAL;
>>
>> I realise we already did the 0x03 check above, but I still think this
>> would be clearer simply as:
>>
>> 	    (addr & 0x3f == 60)
> 
> Indeed, I like the use of `60' there -- hex is overrated ;)

Sure. Will resend.

Ravi
Michael Ellerman March 10, 2021, 5:13 a.m. UTC | #5 
Ravi Bangoria <ravi.bangoria@linux.ibm.com> writes:
> On 3/9/21 4:51 PM, Naveen N. Rao wrote:
>> On 2021/03/09 08:54PM, Michael Ellerman wrote:
>>> Ravi Bangoria <ravi.bangoria@linux.ibm.com> writes:
>>>> As per ISA 3.1, prefixed instruction should not cross 64-byte
>>>> boundary. So don't allow Uprobe on such prefixed instruction.
>>>>
>>>> There are two ways probed instruction is changed in mapped pages.
>>>> First, when Uprobe is activated, it searches for all the relevant
>>>> pages and replace instruction in them. In this case, if that probe
>>>> is on the 64-byte unaligned prefixed instruction, error out
>>>> directly. Second, when Uprobe is already active and user maps a
>>>> relevant page via mmap(), instruction is replaced via mmap() code
>>>> path. But because Uprobe is invalid, entire mmap() operation can
>>>> not be stopped. In this case just print an error and continue.
>>>>
>>>> Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
>>>> Acked-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
>>>
>>> Do we have a Fixes: tag for this?
>> 
>> Since this is an additional check we are adding, I don't think we should
>> add a Fixes: tag. Nothing is broken per-se -- we're just adding more
>> checks to catch simple mistakes. Also, like Oleg pointed out, there are
>> still many other ways for users to shoot themselves in the foot with
>> uprobes and prefixed instructions, if they so desire.
>> 
>> However, if you still think we should add a Fixes: tag, we can perhaps
>> use the below commit since I didn't see any specific commit adding
>> support for prefixed instructions for uprobes:
>> 
>> Fixes: 650b55b707fdfa ("powerpc: Add prefixed instructions to
>> instruction data type")
>
> True. IMO, It doesn't really need any Fixes tag.

Yep OK, I'm happy without a Fixes tag based on that explanation.

>>>> diff --git a/arch/powerpc/kernel/uprobes.c b/arch/powerpc/kernel/uprobes.c
>>>> index e8a63713e655..4cbfff6e94a3 100644
>>>> --- a/arch/powerpc/kernel/uprobes.c
>>>> +++ b/arch/powerpc/kernel/uprobes.c
>>>> @@ -41,6 +41,13 @@ int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe,
>>>>   	if (addr & 0x03)
>>>>   		return -EINVAL;
>>>>   
>>>> +	if (cpu_has_feature(CPU_FTR_ARCH_31) &&
>>>> +	    ppc_inst_prefixed(auprobe->insn) &&
>>>> +	    (addr & (SZ_64 - 4)) == SZ_64 - 4) {
>>>> +		pr_info_ratelimited("Cannot register a uprobe on 64 byte unaligned prefixed instruction\n");
>>>> +		return -EINVAL;
>>>
>>> I realise we already did the 0x03 check above, but I still think this
>>> would be clearer simply as:
>>>
>>> 	    (addr & 0x3f == 60)
>> 
>> Indeed, I like the use of `60' there -- hex is overrated ;)
>
> Sure. Will resend.

Thanks.

cheers
diff mbox series

Patch

diff --git a/arch/powerpc/kernel/uprobes.c b/arch/powerpc/kernel/uprobes.c
index e8a63713e655..4cbfff6e94a3 100644
--- a/arch/powerpc/kernel/uprobes.c
+++ b/arch/powerpc/kernel/uprobes.c
@@ -41,6 +41,13 @@  int arch_uprobe_analyze_insn(struct arch_uprobe *auprobe,
 	if (addr & 0x03)
 		return -EINVAL;
 
+	if (cpu_has_feature(CPU_FTR_ARCH_31) &&
+	    ppc_inst_prefixed(auprobe->insn) &&
+	    (addr & (SZ_64 - 4)) == SZ_64 - 4) {
+		pr_info_ratelimited("Cannot register a uprobe on 64 byte unaligned prefixed instruction\n");
+		return -EINVAL;
+	}
+
 	return 0;
 }