[PATCH] ppc32: fix ppc440 pagetable attributes

[PATCH] ppc32: fix ppc440 pagetable attributes

[Matt Porter]

This patch fixes a bug in the PPC440 pagetable attributes that breaks  
swap support.  It also adds some notes on the PPC440 attribute fields.

Signed-off-by: Geoff Levand <geoffrey.levand@am.sony.com> for CELF
Signed-off-by: Matt Porter <mporter@kernel.crashing.org>

Index: linux-2.6.12-bhpm/include/asm-ppc/pgtable.h
===================================================================
--- linux-2.6.12-bhpm.orig/include/asm-ppc/pgtable.h	2005-06-02 15:09:24.000000000 -0700
+++ linux-2.6.12-bhpm/include/asm-ppc/pgtable.h	2005-06-02 15:47:53.000000000 -0700
@@ -202,18 +202,64 @@
  *
  * Note that these bits preclude future use of a page size
  * less than 4KB.
+ *
+ *
+ * PPC 440 core has following TLB attribute fields;
+ *
+ *   TLB1:
+ *   0  1  2  3  4  ... 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
+ *   RPN.................................  -  -  -  -  -  - ERPN.......
+ *
+ *   TLB2:
+ *   0  1  2  3  4  ... 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
+ *   -  -  -  -  -    - U0 U1 U2 U3 W  I  M  G  E   - UX UW UR SX SW SR
+ *
+ * There are some constrains and options, to decide mapping software bits
+ * into TLB entry.
+ *
+ *   - PRESENT *must* be in the bottom three bits because swap cache
+ *     entries use the top 29 bits for TLB2.
+ *
+ *   - FILE *must* be in the bottom three bits because swap cache
+ *     entries use the top 29 bits for TLB2.
+ *
+ *   - CACHE COHERENT bit (M) has no effect on PPC440 core, because it
+ *     doesn't support SMP. So we can use this as software bit, like
+ *     DIRTY.
+ *
+ * PPC Book-E Linux implementation uses PPC HW PTE bit field definition, 
+ * even it doesn't have HW PTE. 0-11th LSB of PTE stand for memory 
+ * protection-related function. (See PTE structure in include/asm-ppc/mmu.h) 
+ * Definition of _PAGE_XXX in "include/asm-ppc/pagetable.h" stands for 
+ * above bits. Note that those bits values are CPU dependent, not 
+ * architecture.
+ * 
+ * Kernel PTE entry holds arch-dependent swp_entry structure under certain 
+ * situation. In other words, in such situation, some portion of PTE bits 
+ * are used as swp_entry. In PPC implementation, 3-24th LSB are shared with 
+ * swp_entry, however 0-2nd three LSB still hold protection values. 
+ * That means three protection bits are reserved for both PTE and SWAP 
+ * entry at the most three LSBs.
+ *
+ * There are three protection bits available for SWAP entry;
+ *	_PAGE_PRESENT
+ *	_PAGE_FILE
+ *	_PAGE_HASHPTE (if HW has)
+ *
+ * So those three bits have to be inside of 0-2nd LSB of PTE.
+ *
  */
+
 #define _PAGE_PRESENT	0x00000001		/* S: PTE valid */
 #define	_PAGE_RW	0x00000002		/* S: Write permission */
-#define	_PAGE_DIRTY	0x00000004		/* S: Page dirty */
+#define _PAGE_FILE	0x00000004		/* S: nonlinear file mapping */
 #define _PAGE_ACCESSED	0x00000008		/* S: Page referenced */
 #define _PAGE_HWWRITE	0x00000010		/* H: Dirty & RW */
 #define _PAGE_HWEXEC	0x00000020		/* H: Execute permission */
 #define	_PAGE_USER	0x00000040		/* S: User page */
 #define	_PAGE_ENDIAN	0x00000080		/* H: E bit */
 #define	_PAGE_GUARDED	0x00000100		/* H: G bit */
-#define	_PAGE_COHERENT	0x00000200		/* H: M bit */
-#define _PAGE_FILE	0x00000400		/* S: nonlinear file mapping */
+#define	_PAGE_DIRTY	0x00000200		/* S: Page dirty */
 #define	_PAGE_NO_CACHE	0x00000400		/* H: I bit */
 #define	_PAGE_WRITETHRU	0x00000800		/* H: W bit */

Re: [PATCH] ppc32: fix ppc440 pagetable attributes

[Andrew Morton]

Matt Porter <mporter@kernel.crashing.org> wrote:
>
> This patch fixes a bug in the PPC440 pagetable attributes that breaks  
>  swap support.  It also adds some notes on the PPC440 attribute fields.

hm, that looks pretty serious, but it affects all ppc's, yes?

Is this needed for 2.6.13?   Are you super-sure it's safe?

Re: [PATCH] ppc32: fix ppc440 pagetable attributes

[Matt Porter]

On Fri, Aug 05, 2005 at 03:54:42PM -0700, Andrew Morton wrote:
> Matt Porter <mporter@kernel.crashing.org> wrote:
> >
> > This patch fixes a bug in the PPC440 pagetable attributes that breaks  
> >  swap support.  It also adds some notes on the PPC440 attribute fields.
> 
> hm, that looks pretty serious, but it affects all ppc's, yes?

Just PPC440. I originally organized the attributes incorrectly.
 
> Is this needed for 2.6.13?   Are you super-sure it's safe?

This should go into 2.6.13.  I doublechecked it by inspection and
ran numerous tests. It's only existed this long (years) due to the
fact that few people run a disk-based PPC440 system. I would have
sent it up sooner except I wanted to be very sure of my testing
since I don't always trust my own inspection. :)

-Matt