Re: [RFC v5 38/38] Documentation: PowerPC specific updates to memory protection keys

From: Dave Hansen
Date: Tue Jul 11 2017 - 14:23:49 EST


On 07/05/2017 02:22 PM, Ram Pai wrote:
> Add documentation updates that capture PowerPC specific changes.
>
> Signed-off-by: Ram Pai <linuxram@xxxxxxxxxx>
> ---
> Documentation/vm/protection-keys.txt | 85 ++++++++++++++++++++++++++--------
> 1 files changed, 65 insertions(+), 20 deletions(-)
>
> diff --git a/Documentation/vm/protection-keys.txt b/Documentation/vm/protection-keys.txt
> index b643045..d50b6ab 100644
> --- a/Documentation/vm/protection-keys.txt
> +++ b/Documentation/vm/protection-keys.txt
> @@ -1,21 +1,46 @@
> -Memory Protection Keys for Userspace (PKU aka PKEYs) is a CPU feature
> -which will be found on future Intel CPUs.
> +Memory Protection Keys for Userspace (PKU aka PKEYs) is a CPU feature found in
> +new generation of intel CPUs and on PowerPC 7 and higher CPUs.

Please try not to change the wording here. I really did mean to
literally put "future Intel CPUs." Also, you broke my nice wrapping. :)

I'm also thinking that this needs to be more generic. The ppc _CPU_
feature is *NOT* for userspace-only, right?

> Memory Protection Keys provides a mechanism for enforcing page-based
> -protections, but without requiring modification of the page tables
> -when an application changes protection domains. It works by
> -dedicating 4 previously ignored bits in each page table entry to a
> -"protection key", giving 16 possible keys.
> -
> -There is also a new user-accessible register (PKRU) with two separate
> -bits (Access Disable and Write Disable) for each key. Being a CPU
> -register, PKRU is inherently thread-local, potentially giving each
> -thread a different set of protections from every other thread.
> -
> -There are two new instructions (RDPKRU/WRPKRU) for reading and writing
> -to the new register. The feature is only available in 64-bit mode,
> -even though there is theoretically space in the PAE PTEs. These
> -permissions are enforced on data access only and have no effect on
> +protections, but without requiring modification of the page tables when an
> +application changes protection domains.
> +
> +
> +On Intel:
> +
> + It works by dedicating 4 previously ignored bits in each page table
> + entry to a "protection key", giving 16 possible keys.
> +
> + There is also a new user-accessible register (PKRU) with two separate
> + bits (Access Disable and Write Disable) for each key. Being a CPU
> + register, PKRU is inherently thread-local, potentially giving each
> + thread a different set of protections from every other thread.
> +
> + There are two new instructions (RDPKRU/WRPKRU) for reading and writing
> + to the new register. The feature is only available in 64-bit mode,
> + even though there is theoretically space in the PAE PTEs. These
> + permissions are enforced on data access only and have no effect on
> + instruction fetches.
> +
> +
> +On PowerPC:
> +
> + It works by dedicating 5 page table entry bits to a "protection key",
> + giving 32 possible keys.
> +
> + There is a user-accessible register (AMR) with two separate bits;
> + Access Disable and Write Disable, for each key. Being a CPU
> + register, AMR is inherently thread-local, potentially giving each
> + thread a different set of protections from every other thread. NOTE:
> + Disabling read permission does not disable write and vice-versa.
> +
> + The feature is available on 64-bit HPTE mode only.
> + 'mtspr 0xd, mem' reads the AMR register
> + 'mfspr mem, 0xd' writes into the AMR register.

The whole "being a CPU register" bits seem pretty common. Should it be
in the leading paragraph that is shared?

> +Permissions are enforced on data access only and have no effect on
> instruction fetches.

Shouldn't we mention the ppc support for execute-disable here too?

Also, *does* this apply to ppc? You have it both in this common area
and in the x86 portion.

> =========================== Syscalls ===========================
> @@ -28,9 +53,9 @@ There are 3 system calls which directly interact with pkeys:
> unsigned long prot, int pkey);
>
> Before a pkey can be used, it must first be allocated with
> -pkey_alloc(). An application calls the WRPKRU instruction
> +pkey_alloc(). An application calls the WRPKRU/AMR instruction
> directly in order to change access permissions to memory covered
> -with a key. In this example WRPKRU is wrapped by a C function
> +with a key. In this example WRPKRU/AMR is wrapped by a C function
> called pkey_set().
>
> int real_prot = PROT_READ|PROT_WRITE;
> @@ -52,11 +77,11 @@ is no longer in use:
> munmap(ptr, PAGE_SIZE);
> pkey_free(pkey);
>
> -(Note: pkey_set() is a wrapper for the RDPKRU and WRPKRU instructions.
> +(Note: pkey_set() is a wrapper for the RDPKRU,WRPKRU or AMR instructions.
> An example implementation can be found in
> tools/testing/selftests/x86/protection_keys.c)
>
> -=========================== Behavior ===========================
> +=========================== Behavior =================================
>
> The kernel attempts to make protection keys consistent with the
> behavior of a plain mprotect(). For instance if you do this:
> @@ -83,3 +108,23 @@ with a read():
> The kernel will send a SIGSEGV in both cases, but si_code will be set
> to SEGV_PKERR when violating protection keys versus SEGV_ACCERR when
> the plain mprotect() permissions are violated.
> +
> +
> +====================================================================
> + Semantic differences
> +
> +The following semantic differences exist between x86 and power.
> +
> +a) powerpc allows creation of a key with execute-disabled. The following
> + is allowed on powerpc.
> + pkey = pkey_alloc(0, PKEY_DISABLE_WRITE | PKEY_DISABLE_ACCESS |
> + PKEY_DISABLE_EXECUTE);
> + x86 disallows PKEY_DISABLE_EXECUTE during key creation.

It isn't that powerpc supports *creation* of the key. It doesn't
support setting PKEY_DISABLE_EXECUTE, period, which implies that you
can't set it at pkey_alloc(). That's a pretty important distinction, IMNHO.

> +b) changing the permission bits of a key from a signal handler does not
> + persist on x86. The PKRU specific fpregs entry needs to be modified
> + for it to persist. On powerpc the permission bits of the key can be
> + modified by programming the AMR register from the signal handler.
> + The changes persists across signal boundaries.

^"changes persist", not "persists".