Message-Id: <20170621000436.GN17588@ram.oc3035372033.ibm.com>
Date:   Tue, 20 Jun 2017 17:04:36 -0700
From:   Ram Pai <linuxram@...ibm.com>
To:     Anshuman Khandual <khandual@...ux.vnet.ibm.com>
Cc:     linuxppc-dev@...ts.ozlabs.org, linux-kernel@...r.kernel.org,
        benh@...nel.crashing.org, paulus@...ba.org, mpe@...erman.id.au,
        aneesh.kumar@...ux.vnet.ibm.com, bsingharora@...il.com,
        dave.hansen@...el.com, hbabu@...ibm.com
Subject: Re: [RFC v2 11/12]Documentation: Documentation updates.

On Tue, Jun 20, 2017 at 11:48:23AM +0530, Anshuman Khandual wrote:
> On 06/17/2017 09:22 AM, Ram Pai wrote:
> > The Documentaton file is moved from x86 into the generic area,
> > since this feature is now supported by more than one archs.
> > 
> > Signed-off-by: Ram Pai <linuxram@...ibm.com>
> > ---
> >  Documentation/vm/protection-keys.txt  | 110 ++++++++++++++++++++++++++++++++++
> >  Documentation/x86/protection-keys.txt |  85 --------------------------
> 
> I am not sure whether this is a good idea. There might be
> specifics for each architecture which need to be detailed
> again in this new generic one.
> 
> >  2 files changed, 110 insertions(+), 85 deletions(-)
> >  create mode 100644 Documentation/vm/protection-keys.txt
> >  delete mode 100644 Documentation/x86/protection-keys.txt
> > 
> > diff --git a/Documentation/vm/protection-keys.txt b/Documentation/vm/protection-keys.txt
> > new file mode 100644
> > index 0000000..b49e6bb
> > --- /dev/null
> > +++ b/Documentation/vm/protection-keys.txt
> > @@ -0,0 +1,110 @@
> > +Memory Protection Keys for Userspace (PKU aka PKEYs) is a CPU feature
> > +found in new generation of intel CPUs on PowerPC CPUs.
> > +
> > +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.
> 
> Does resultant access through protection keys should be a
> subset of the protection bits enabled through original PTE
> PROT format ? Does the semantics exactly the same on x86
> and powerpc ?

The protection key takes precedence over protection done through
mprotect. 
Yes both on x86 and powerpc we maintain the same semantics.
> 
> > +
> > +
> > +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 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.
> 
> Small nit. Space needed here.
> 
> > +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.
> > +
> > +Permissions are enforced on data access only and have no effect on
> > +instruction fetches.
> > +
> > +=========================== Syscalls ===========================
> > +
> > +There are 3 system calls which directly interact with pkeys:
> > +
> > +	int pkey_alloc(unsigned long flags, unsigned long init_access_rights)
> > +	int pkey_free(int pkey);
> > +	int pkey_mprotect(unsigned long start, size_t len,
> > +			  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
> > +directly in order to change access permissions to memory covered
> > +with a key.  In this example WRPKRU is wrapped by a C function
> > +called pkey_set().
> > +
> > +	int real_prot = PROT_READ|PROT_WRITE;
> > +	pkey = pkey_alloc(0, PKEY_DENY_WRITE);
> > +	ptr = mmap(NULL, PAGE_SIZE, PROT_NONE, MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
> > +	ret = pkey_mprotect(ptr, PAGE_SIZE, real_prot, pkey);
> > +	... application runs here
> > +
> > +Now, if the application needs to update the data at 'ptr', it can
> > +gain access, do the update, then remove its write access:
> > +
> > +	pkey_set(pkey, 0); // clear PKEY_DENY_WRITE
> > +	*ptr = foo; // assign something
> > +	pkey_set(pkey, PKEY_DENY_WRITE); // set PKEY_DENY_WRITE again
> > +
> > +Now when it frees the memory, it will also free the pkey since it
> > +is no longer in use:
> > +
> > +	munmap(ptr, PAGE_SIZE);
> > +	pkey_free(pkey);
> > +
> > +(Note: pkey_set() is a wrapper for the RDPKRU and WRPKRU instructions.
> > + An example implementation can be found in
> > + tools/testing/selftests/x86/protection_keys.c)
> > +
> > +=========================== Behavior ===========================
> > +
> > +The kernel attempts to make protection keys consistent with the
> > +behavior of a plain mprotect().  For instance if you do this:
> > +
> > +	mprotect(ptr, size, PROT_NONE);
> > +	something(ptr);
> > +
> > +you can expect the same effects with protection keys when doing this:
> > +
> > +	pkey = pkey_alloc(0, PKEY_DISABLE_WRITE | PKEY_DISABLE_READ);
> > +	pkey_mprotect(ptr, size, PROT_READ|PROT_WRITE, pkey);
> > +	something(ptr);
> > +
> > +That should be true whether something() is a direct access to 'ptr'
> > +like:
> > +
> > +	*ptr = foo;
> > +
> > +or when the kernel does the access on the application's behalf like
> > +with a read():
> > +
> > +	read(fd, ptr, 1);
> > +
> > +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.
> 
> I guess the right thing would be to have three files
> 
> * Documentation/vm/protection-keys.txt
> 
> 	- Generic interface, system calls
> 	- Signal handling, error codes
> 	- Semantics of programming with an example
> 
> * Documentation/x86/protection-keys.txt
> 
> 	- Number of active protections keys inside an address space
> 	- X86 protection key instruction details
> 	- PTE protection bits placement details
> 	- Page fault handling
> 	- Implementation details a bit ?
> 
> * Documentation/powerpc/protection-keys.txt
> 
> 	- Number of active protections keys inside an address space
> 	- Powerpc instructions details
> 	- PTE protection bits placement details
> 	- Page fault handling
> 	- Implementation details a bit ?

I see the value of your suggestion. This is something that will touch
atleast two architectures. Want to hear some more inputs before I do the
changes.  

Dave Hansen:  would like to hear your ideas.

RP

Powered by blists - more mailing lists