[PATCH] Fix PPC440 pagetable attributes
Geoff Levand
geoffrey.levand at am.sony.com
Sat Jun 4 02:30:37 EST 2005
Kumar Gala wrote:
> On Jun 2, 2005, at 6:00 PM, Geoff Levand 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.
>>
>> *
>> * 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.
>>+ *
>
> I disagree with this comment. PPC Book-E PTE format has nothing to do
> with PPC HW PTE format.
>
OK, is this more agreeable?
* With the PPC Book-E Linux implementation, 0-11th LSB of PTE stand for memory
* protection-related function. (See PTE structure in include/asm-ppc/mmu.h)
* Definition of _PAGE_XXX here stands for above bits. Note that those bits
* values are CPU dependent, not architecture.
If not, could you be more specific.
-Geoff
More information about the Linuxppc-embedded
mailing list