[PATCH v3 1/2] powerpc: fix KUAP warning in VMX usercopy path

Sayali Patil sayalip at linux.ibm.com
Wed Mar 4 23:30:48 AEDT 2026


On 04/03/26 12:19, Christophe Leroy (CS GROUP) wrote:
> Hi Sayali,
>
> Le 04/03/2026 à 06:35, Sayali Patil a écrit :
>> On powerpc with PREEMPT_FULL or PREEMPT_LAZY and function tracing 
>> enabled,
>> KUAP warnings can be triggered from the VMX usercopy path under memory
>> stress workloads.
>>
>> KUAP requires that no subfunctions are called once userspace access has
>> been enabled. The existing VMX copy implementation violates this
>> requirement by invoking enter_vmx_usercopy() from the assembly path 
>> after
>> userspace access has already been enabled. If preemption occurs
>> in this window, the AMR state may not be preserved correctly,
>> leading to unexpected userspace access state and resulting in
>> KUAP warnings.
>>
>> Fix this by restructuring the VMX usercopy flow so that VMX selection
>> and VMX state management are centralized in raw_copy_tofrom_user(),
>> which is invoked by the raw_copy_{to,from,in}_user() wrappers.
>>
>> The new flow is:
>>
>>    - raw_copy_{to,from,in}_user() calls raw_copy_tofrom_user()
>>    - raw_copy_tofrom_user() decides whether to use the VMX path
>>      based on size and CPU capability
>>    - Call enter_vmx_usercopy() before enabling userspace access
>>    - Enable userspace access as per the copy direction
>>      and perform the VMX copy
>>    - Disable userspace access as per the copy direction
>>    - Call exit_vmx_usercopy()
>>    - Fall back to the base copy routine if the VMX copy faults
>>
>> With this change, the VMX assembly routines no longer perform VMX state
>> management or call helper functions; they only implement the
>> copy operations.
>> The previous feature-section based VMX selection inside
>> __copy_tofrom_user_power7() is removed, and a dedicated
>> __copy_tofrom_user_power7_vmx() entry point is introduced.
>>
>> This ensures correct KUAP ordering, avoids subfunction calls
>> while KUAP is unlocked, and eliminates the warnings while preserving
>> the VMX fast path.
>>
>> Fixes: de78a9c42a79 ("powerpc: Add a framework for Kernel Userspace 
>> Access Protection")
>> Reported-by: Shrikanth Hegde <sshegde at linux.ibm.com>
>> Closes: 
>> https://lore.kernel.org/all/20260109064917.777587-2-sshegde@linux.ibm.com/
>> Suggested-by: Christophe Leroy <chleroy at kernel.org>
>> Co-developed-by: Aboorva Devarajan <aboorvad at linux.ibm.com>
>> Signed-off-by: Aboorva Devarajan <aboorvad at linux.ibm.com>
>> Signed-off-by: Sayali Patil <sayalip at linux.ibm.com>
>
> That looks almost good, some editorial comments below.
>
> With those fixed, you can add  Reviewed-by: Christophe Leroy (CS 
> GROUP) <chleroy at kernel.org>
>
>> ---
>>
>> v2->v3
>>    - Addressd as per review feedback by removing usercopy_mode enum
>>      and using the copy direction directly for KUAP permission handling.
>>    - Integrated __copy_tofrom_user_vmx() functionality into
>>      raw_copy_tofrom_user() in uaccess.h as a static __always_inline
>>      implementation.
>>    - Exported enter_vmx_usercopy() and exit_vmx_usercopy()
>>      to support VMX usercopy handling from the common path.
>>
>> v2: 
>> https://lore.kernel.org/all/20260228135319.238985-1-sayalip@linux.ibm.com/
>>
>> ---
>>   arch/powerpc/include/asm/uaccess.h | 66 ++++++++++++++++++++++--------
>>   arch/powerpc/lib/copyuser_64.S     |  1 +
>>   arch/powerpc/lib/copyuser_power7.S | 45 +++++++-------------
>>   arch/powerpc/lib/vmx-helper.c      |  2 +
>>   4 files changed, 66 insertions(+), 48 deletions(-)
>>
>> diff --git a/arch/powerpc/include/asm/uaccess.h 
>> b/arch/powerpc/include/asm/uaccess.h
>> index ba1d878c3f40..8fd412671025 100644
>> --- a/arch/powerpc/include/asm/uaccess.h
>> +++ b/arch/powerpc/include/asm/uaccess.h
>> @@ -15,6 +15,9 @@
>>   #define TASK_SIZE_MAX        TASK_SIZE_USER64
>>   #endif
>>   +/* Threshold above which VMX copy path is used */
>> +#define VMX_COPY_THRESHOLD 3328
>> +
>>   #include <asm-generic/access_ok.h>
>>     /*
>> @@ -326,40 +329,67 @@ do {                                \
>>   extern unsigned long __copy_tofrom_user(void __user *to,
>>           const void __user *from, unsigned long size);
>>   -#ifdef __powerpc64__
>> -static inline unsigned long
>> -raw_copy_in_user(void __user *to, const void __user *from, unsigned 
>> long n)
>> +unsigned long __copy_tofrom_user_base(void __user *to,
>> +        const void __user *from, unsigned long size);
>> +
>> +unsigned long __copy_tofrom_user_power7_vmx(void __user *to,
>> +        const void __user *from, unsigned long size);
>> +
>> +
>
> Remove one line.
>
>> +static __always_inline bool will_use_vmx(unsigned long n)
>> +{
>> +    return IS_ENABLED(CONFIG_ALTIVEC) &&
>> +        cpu_has_feature(CPU_FTR_VMX_COPY) &&
>> +        n > VMX_COPY_THRESHOLD;
>
> Avoid too many line when possible. Nowadays up to 100 chars per line 
> are allowed.
>
> Take care of alignment of second line, the second line should start at 
> same position as IS_ENABLED, meaning you have to insert 7 spaces 
> instead of a tab.
>
>> +}
>> +
>> +static __always_inline unsigned long raw_copy_tofrom_user(void 
>> __user *to,
>> +        const void __user *from, unsigned long n,
>> +        unsigned long dir)
>
> Subsequent lines should start at same position as the ( of the first 
> line, therefore I'd suggest following form instead:
>
> static __always_inline unsigned long
> raw_copy_tofrom_user(void __user *to,const void __user *from, unsigned 
> long n, unsigned long dir)
>
>>   {
>>       unsigned long ret;
>>   -    barrier_nospec();
>> -    allow_user_access(to, KUAP_READ_WRITE);
>> +    if (will_use_vmx(n) && enter_vmx_usercopy()) {
>> +        allow_user_access(to, dir);
>> +        ret = __copy_tofrom_user_power7_vmx(to, from, n);
>> +        prevent_user_access(dir);
>> +        exit_vmx_usercopy();
>> +
>> +        if (unlikely(ret)) {
>> +            allow_user_access(to, dir);
>> +            ret = __copy_tofrom_user_base(to, from, n);
>> +            prevent_user_access(dir);
>> +        }
>> +        return ret;
>> +    }
>> +
>> +    allow_user_access(to, dir);
>>       ret = __copy_tofrom_user(to, from, n);
>> -    prevent_user_access(KUAP_READ_WRITE);
>> +    prevent_user_access(dir);
>>       return ret;
>>   }
>> +
>> +#ifdef __powerpc64__
>
> I know it was already there before, but checkpatch is not happy about 
> __power64__. It should be replaced by CONFIG_PPC64.
>
>> +static inline unsigned long
>> +raw_copy_in_user(void __user *to, const void __user *from, unsigned 
>> long n)
>> +{
>> +    barrier_nospec();
>> +    return raw_copy_tofrom_user(to, from, n, KUAP_READ_WRITE);
>> +}
>>   #endif /* __powerpc64__ */
>>     static inline unsigned long raw_copy_from_user(void *to,
>>           const void __user *from, unsigned long n)
>
> Same problem with alignment of second line. Prefer the form used for 
> raw_copy_in_user() or raw_copy_to_user(), ie:
>
> static inline unsigned long
> raw_copy_from_user(void *to, const void __user *from, unsigned long n)
>
>>   {
>> -    unsigned long ret;
>> -
>> -    allow_user_access(NULL, KUAP_READ);
>> -    ret = __copy_tofrom_user((__force void __user *)to, from, n);
>> -    prevent_user_access(KUAP_READ);
>> -    return ret;
>> +    return raw_copy_tofrom_user((__force void __user *)to, from,
>> +                    n, KUAP_READ);
>
> 100 chars are allowed per line, this should fit on a single line.
>
>>   }
>>     static inline unsigned long
>>   raw_copy_to_user(void __user *to, const void *from, unsigned long n)
>>   {
>> -    unsigned long ret;
>> -
>> -    allow_user_access(to, KUAP_WRITE);
>> -    ret = __copy_tofrom_user(to, (__force const void __user *)from, n);
>> -    prevent_user_access(KUAP_WRITE);
>> -    return ret;
>> +    return raw_copy_tofrom_user(to, (__force const void __user *)from,
>> +                    n, KUAP_WRITE);
>
> 100 chars are allowed per line, this should fit on a single line.
>
>>   }
>>     unsigned long __arch_clear_user(void __user *addr, unsigned long 
>> size);
>
>
> Run checkpatch before submitting patches:
>
> $ ./scripts/checkpatch.pl --strict -g HEAD~
> CHECK: Alignment should match open parenthesis
> #83: FILE: arch/powerpc/include/asm/uaccess.h:333:
> +unsigned long __copy_tofrom_user_base(void __user *to,
> +        const void __user *from, unsigned long size);
>
> CHECK: Alignment should match open parenthesis
> #86: FILE: arch/powerpc/include/asm/uaccess.h:336:
> +unsigned long __copy_tofrom_user_power7_vmx(void __user *to,
> +        const void __user *from, unsigned long size);
>
> CHECK: Please don't use multiple blank lines
> #88: FILE: arch/powerpc/include/asm/uaccess.h:338:
> +
> +
>
> CHECK: Alignment should match open parenthesis
> #97: FILE: arch/powerpc/include/asm/uaccess.h:347:
> +static __always_inline unsigned long raw_copy_tofrom_user(void __user 
> *to,
> +        const void __user *from, unsigned long n,
>
> CHECK: architecture specific defines should be avoided
> #125: FILE: arch/powerpc/include/asm/uaccess.h:372:
> +#ifdef __powerpc64__
>
> total: 0 errors, 0 warnings, 5 checks, 212 lines checked
>
> NOTE: For some of the reported defects, checkpatch may be able to
>       mechanically convert to the typical style using --fix or 
> --fix-inplace.
>
> Commit 3a44f6614d88 ("powerpc: fix KUAP warning in VMX usercopy path") 
> has style problems, please review.
>
> NOTE: If any of the errors are false positives, please report
>       them to the maintainer, see CHECKPATCH in MAINTAINERS.
>
Thanks Christophe for the review.
I have addressed the comments and incorporated the changes in v4.

As suggested, I have added:
Reviewed-by: Christophe Leroy (CS GROUP) <chleroy at kernel.org>

v4: 
https://lore.kernel.org/all/20260304122201.153049-1-sayalip@linux.ibm.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ozlabs.org/pipermail/linuxppc-dev/attachments/20260304/7db1d9b5/attachment.htm>


More information about the Linuxppc-dev mailing list