Re: [PATCH v4 1/4] pkeys: add API to switch to permissive/zero pkey register

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

On 2025-02-24 08:20, Dmitry Vyukov wrote:
> The API allows to switch to permissive pkey register that allows accesses
> to all PKEYs, and to a value that allows acccess to the 0 (default) PKEY.
> This functionality is already used in x86 signal delivery, and will be
> needed for rseq.

AFAIU the signal delivery uses the "allow access to all PKEYs"
approach only, not the "allow access to 0-pkey". It would be
good to clarify this in the commit message.

> 
> Signed-off-by: Dmitry Vyukov <dvyukov@google.com>
> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: "Paul E. McKenney" <paulmck@kernel.org>
> Cc: Boqun Feng <boqun.feng@gmail.com>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: Borislav Petkov <bp@alien8.de>
> Cc: Dave Hansen <dave.hansen@linux.intel.com>
> Cc: "H. Peter Anvin" <hpa@zytor.com>
> Cc: Aruna Ramakrishna <aruna.ramakrishna@oracle.com>
> Cc: x86@kernel.org
> Cc: linux-kernel@vger.kernel.org
> Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
> Fixes: d7822b1e24f2 ("rseq: Introduce restartable sequences system call")
> 
> ---
> Changes in v4:
>   - Added Fixes tag
> 
> Changes in v3:
>   - Renamed API functions to write_permissive_pkey_val/write_pkey_val
>   - Added enable_zero_pkey_val for rseq
>   - Added Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
> 
> Changes in v2:
>   - Fixed typo in commit description
> ---
>   arch/x86/Kconfig             |  1 +
>   arch/x86/include/asm/pkeys.h | 33 +++++++++++++++++++++++++++++++++
>   arch/x86/include/asm/pkru.h  | 10 +++++++---
>   include/linux/pkeys.h        | 31 +++++++++++++++++++++++++++++++
>   mm/Kconfig                   |  2 ++
>   5 files changed, 74 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> index be2c311f5118d..43af2840d098f 100644
> --- a/arch/x86/Kconfig
> +++ b/arch/x86/Kconfig
> @@ -1881,6 +1881,7 @@ config X86_INTEL_MEMORY_PROTECTION_KEYS
>   	depends on X86_64 && (CPU_SUP_INTEL || CPU_SUP_AMD)
>   	select ARCH_USES_HIGH_VMA_FLAGS
>   	select ARCH_HAS_PKEYS
> +	select ARCH_HAS_PERMISSIVE_PKEY
>   	help
>   	  Memory Protection Keys provides a mechanism for enforcing
>   	  page-based protections, but without requiring modification of the
> diff --git a/arch/x86/include/asm/pkeys.h b/arch/x86/include/asm/pkeys.h
> index 2e6c04d8a45b4..d6e35ab5c3d22 100644
> --- a/arch/x86/include/asm/pkeys.h
> +++ b/arch/x86/include/asm/pkeys.h
> @@ -2,6 +2,8 @@
>   #ifndef _ASM_X86_PKEYS_H
>   #define _ASM_X86_PKEYS_H
>   
> +#include "pkru.h"
> +
>   /*
>    * If more than 16 keys are ever supported, a thorough audit
>    * will be necessary to ensure that the types that store key
> @@ -123,4 +125,35 @@ static inline int vma_pkey(struct vm_area_struct *vma)
>   	return (vma->vm_flags & vma_pkey_mask) >> VM_PKEY_SHIFT;
>   }
>   
> +typedef u32 pkey_reg_t;
> +
> +static inline pkey_reg_t write_permissive_pkey_val(void)
> +{
> +	return write_pkru(0);
> +}
> +
> +static inline pkey_reg_t enable_zero_pkey_val(void)
> +{
> +	u32 pkru;
> +
> +	if (!cpu_feature_enabled(X86_FEATURE_OSPKE))
> +		return 0;
> +	/*
> +	 * WRPKRU is relatively expensive compared to RDPKRU,
> +	 * avoid it if possible.
> +	 */
> +	pkru = rdpkru();
> +	if ((pkru & (PKRU_AD_BIT|PKRU_WD_BIT)) != 0)
> +		wrpkru(pkru & ~(PKRU_AD_BIT|PKRU_WD_BIT));
> +	return pkru;
> +
> +
> +	return write_pkru(0);

This is dead code. What I am missing ?

With those fixed, please keep my reviewed-by :)

Thanks,

Mathieu

> +}
> +
> +static inline void write_pkey_val(pkey_reg_t val)
> +{
> +	write_pkru(val);
> +}
> +
>   #endif /*_ASM_X86_PKEYS_H */
> diff --git a/arch/x86/include/asm/pkru.h b/arch/x86/include/asm/pkru.h
> index 74f0a2d34ffdd..b9bf9b7f2753b 100644
> --- a/arch/x86/include/asm/pkru.h
> +++ b/arch/x86/include/asm/pkru.h
> @@ -39,16 +39,20 @@ static inline u32 read_pkru(void)
>   	return 0;
>   }
>   
> -static inline void write_pkru(u32 pkru)
> +static inline u32 write_pkru(u32 pkru)
>   {
> +	u32 old_pkru;
> +
>   	if (!cpu_feature_enabled(X86_FEATURE_OSPKE))
> -		return;
> +		return 0;
>   	/*
>   	 * WRPKRU is relatively expensive compared to RDPKRU.
>   	 * Avoid WRPKRU when it would not change the value.
>   	 */
> -	if (pkru != rdpkru())
> +	old_pkru = rdpkru();
> +	if (pkru != old_pkru)
>   		wrpkru(pkru);
> +	return old_pkru;
>   }
>   
>   static inline void pkru_write_default(void)
> diff --git a/include/linux/pkeys.h b/include/linux/pkeys.h
> index 86be8bf27b41b..262d60f6a15f8 100644
> --- a/include/linux/pkeys.h
> +++ b/include/linux/pkeys.h
> @@ -48,4 +48,35 @@ static inline bool arch_pkeys_enabled(void)
>   
>   #endif /* ! CONFIG_ARCH_HAS_PKEYS */
>   
> +#ifndef CONFIG_ARCH_HAS_PERMISSIVE_PKEY
> +
> +/*
> + * Common name for value of the register that controls access to PKEYs
> + * (called differently on different arches: PKRU, POR, AMR).
> + */
> +typedef char pkey_reg_t;
> +
> +/*
> + * Sets PKEY access register to the most permissive value that allows
> + * accesses to all PKEYs. Returns the current value of PKEY register.
> + * Code should generally arrange switching back to the old value
> + * using write_pkey_val(old_value).
> + */
> +static inline pkey_reg_t write_permissive_pkey_val(void)
> +{
> +	return 0;
> +}
> +
> +/*
> + * Sets PKEY access register to a value that allows access to the 0 (default)
> + * PKEY. Returns the current value of PKEY register.
> + */
> +static inline pkey_reg_t enable_zero_pkey_val(void)
> +{
> +	return 0;
> +}
> +
> +static inline void write_pkey_val(pkey_reg_t val) {}
> +#endif /* ! CONFIG_ARCH_HAS_PERMISSIVE_PKEY */
> +
>   #endif /* _LINUX_PKEYS_H */
> diff --git a/mm/Kconfig b/mm/Kconfig
> index 1b501db064172..9e874f7713a2b 100644
> --- a/mm/Kconfig
> +++ b/mm/Kconfig
> @@ -1147,6 +1147,8 @@ config ARCH_USES_HIGH_VMA_FLAGS
>   	bool
>   config ARCH_HAS_PKEYS
>   	bool
> +config ARCH_HAS_PERMISSIVE_PKEY
> +	bool
>   
>   config ARCH_USES_PG_ARCH_2
>   	bool


-- 
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com