summaryrefslogtreecommitdiff
path: root/arch/sparc/include/asm/tsb.h
diff options
context:
space:
mode:
authorDavid S. Miller <davem@davemloft.net>2013-09-27 00:45:15 +0400
committerDavid S. Miller <davem@davemloft.net>2013-11-14 00:33:08 +0400
commita7b9403f0e6d5f99139dca18be885819c8d380a1 (patch)
tree3e88ddc9250a1408eefc48a20a8fb445e19fcaa3 /arch/sparc/include/asm/tsb.h
parent2b77933c28f5044629bb19e8045aae65b72b939d (diff)
downloadlinux-a7b9403f0e6d5f99139dca18be885819c8d380a1.tar.xz
sparc64: Encode huge PMDs using PTE encoding.
Now that we have 64-bits for PMDs we can stop using special encodings for the huge PMD values, and just put real PTEs in there. We allocate a _PAGE_PMD_HUGE bit to distinguish between plain PMDs and huge ones. It is the same for both 4U and 4V PTE layouts. We also use _PAGE_SPECIAL to indicate the splitting state, since a huge PMD cannot also be special. All of the PMD --> PTE translation code disappears, and most of the huge PMD bit modifications and tests just degenerate into the PTE operations. In particular USER_PGTABLE_CHECK_PMD_HUGE becomes trivial. As a side effect, normal PMDs don't shift the physical address around. This also speeds up the page table walks in the TLB miss paths since they don't have to do the shifts any more. Another non-trivial aspect is that pte_modify() has to be changed to preserve the _PAGE_PMD_HUGE bits as well as the page size field of the pte. Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'arch/sparc/include/asm/tsb.h')
-rw-r--r--arch/sparc/include/asm/tsb.h92
1 files changed, 12 insertions, 80 deletions
diff --git a/arch/sparc/include/asm/tsb.h b/arch/sparc/include/asm/tsb.h
index cc0432f15817..2230f80d9fe3 100644
--- a/arch/sparc/include/asm/tsb.h
+++ b/arch/sparc/include/asm/tsb.h
@@ -147,100 +147,34 @@ extern struct tsb_phys_patch_entry __tsb_phys_patch, __tsb_phys_patch_end;
brz,pn REG1, FAIL_LABEL; \
sllx VADDR, 64 - (PMD_SHIFT + PMD_BITS), REG2; \
srlx REG2, 64 - PAGE_SHIFT, REG2; \
- sllx REG1, PGD_PADDR_SHIFT, REG1; \
andn REG2, 0x7, REG2; \
ldxa [REG1 + REG2] ASI_PHYS_USE_EC, REG1; \
brz,pn REG1, FAIL_LABEL; \
sllx VADDR, 64 - PMD_SHIFT, REG2; \
srlx REG2, 64 - PAGE_SHIFT, REG2; \
- sllx REG1, PMD_PADDR_SHIFT, REG1; \
andn REG2, 0x7, REG2; \
add REG1, REG2, REG1;
- /* These macros exists only to make the PMD translator below
- * easier to read. It hides the ELF section switch for the
- * sun4v code patching.
- */
-#define OR_PTE_BIT_1INSN(REG, NAME) \
-661: or REG, _PAGE_##NAME##_4U, REG; \
- .section .sun4v_1insn_patch, "ax"; \
- .word 661b; \
- or REG, _PAGE_##NAME##_4V, REG; \
- .previous;
-
-#define OR_PTE_BIT_2INSN(REG, TMP, NAME) \
-661: sethi %hi(_PAGE_##NAME##_4U), TMP; \
- or REG, TMP, REG; \
- .section .sun4v_2insn_patch, "ax"; \
- .word 661b; \
- mov -1, TMP; \
- or REG, _PAGE_##NAME##_4V, REG; \
- .previous;
-
- /* Load into REG the PTE value for VALID, CACHE, and SZHUGE.
- *
- * We are fabricating an 8MB page using 2 4MB HW pages here.
- */
-#define BUILD_PTE_VALID_SZHUGE_CACHE(VADDR, PADDR_BITS, REG) \
- sethi %hi(4 * 1024 * 1024), REG; \
- andn PADDR_BITS, REG, PADDR_BITS; \
- and VADDR, REG, REG; \
- or PADDR_BITS, REG, PADDR_BITS; \
-661: sethi %uhi(_PAGE_VALID|_PAGE_SZHUGE_4U), REG; \
- .section .sun4v_1insn_patch, "ax"; \
- .word 661b; \
- sethi %uhi(_PAGE_VALID), REG; \
- .previous; \
- sllx REG, 32, REG; \
-661: or REG, _PAGE_CP_4U|_PAGE_CV_4U, REG; \
- .section .sun4v_1insn_patch, "ax"; \
- .word 661b; \
- or REG, _PAGE_CP_4V|_PAGE_CV_4V|_PAGE_SZHUGE_4V, REG; \
- .previous;
-
/* PMD has been loaded into REG1, interpret the value, seeing
* if it is a HUGE PMD or a normal one. If it is not valid
* then jump to FAIL_LABEL. If it is a HUGE PMD, and it
* translates to a valid PTE, branch to PTE_LABEL.
*
- * We translate the PMD by hand, one bit at a time,
- * constructing the huge PTE.
- *
- * So we construct the PTE in REG2 as follows:
- *
- * 1) Extract the PMD PFN from REG1 and place it into REG2.
- *
- * 2) Translate PMD protection bits in REG1 into REG2, one bit
- * at a time using andcc tests on REG1 and OR's into REG2.
- *
- * Only two bits to be concerned with here, EXEC and WRITE.
- * Now REG1 is freed up and we can use it as a temporary.
- *
- * 3) Construct the VALID, CACHE, and page size PTE bits in
- * REG1, OR with REG2 to form final PTE.
+ * We have to propagate the 4MB bit of the virtual address
+ * because we are fabricating 8MB pages using 4MB hw pages.
*/
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
#define USER_PGTABLE_CHECK_PMD_HUGE(VADDR, REG1, REG2, FAIL_LABEL, PTE_LABEL) \
- brz,pn REG1, FAIL_LABEL; \
- andcc REG1, PMD_ISHUGE, %g0; \
- be,pt %xcc, 700f; \
- and REG1, PMD_HUGE_PRESENT|PMD_HUGE_ACCESSED, REG2; \
- cmp REG2, PMD_HUGE_PRESENT|PMD_HUGE_ACCESSED; \
- bne,pn %xcc, FAIL_LABEL; \
- andn REG1, PMD_HUGE_PROTBITS, REG2; \
- sllx REG2, PMD_PADDR_SHIFT, REG2; \
- /* REG2 now holds PFN << PAGE_SHIFT */ \
- andcc REG1, PMD_HUGE_WRITE, %g0; \
- bne,a,pt %xcc, 1f; \
- OR_PTE_BIT_1INSN(REG2, W); \
-1: andcc REG1, PMD_HUGE_EXEC, %g0; \
- be,pt %xcc, 1f; \
- nop; \
- OR_PTE_BIT_2INSN(REG2, REG1, EXEC); \
- /* REG1 can now be clobbered, build final PTE */ \
-1: BUILD_PTE_VALID_SZHUGE_CACHE(VADDR, REG2, REG1); \
- ba,pt %xcc, PTE_LABEL; \
- or REG1, REG2, REG1; \
+ brz,pn REG1, FAIL_LABEL; \
+ sethi %uhi(_PAGE_PMD_HUGE), REG2; \
+ sllx REG2, 32, REG2; \
+ andcc REG1, REG2, %g0; \
+ be,pt %xcc, 700f; \
+ sethi %hi(4 * 1024 * 1024), REG2; \
+ andn REG1, REG2, REG1; \
+ and VADDR, REG2, REG2; \
+ brlz,pt REG1, PTE_LABEL; \
+ or REG1, REG2, REG1; \
700:
#else
#define USER_PGTABLE_CHECK_PMD_HUGE(VADDR, REG1, REG2, FAIL_LABEL, PTE_LABEL) \
@@ -265,13 +199,11 @@ extern struct tsb_phys_patch_entry __tsb_phys_patch, __tsb_phys_patch_end;
brz,pn REG1, FAIL_LABEL; \
sllx VADDR, 64 - (PMD_SHIFT + PMD_BITS), REG2; \
srlx REG2, 64 - PAGE_SHIFT, REG2; \
- sllx REG1, PGD_PADDR_SHIFT, REG1; \
andn REG2, 0x7, REG2; \
ldxa [REG1 + REG2] ASI_PHYS_USE_EC, REG1; \
USER_PGTABLE_CHECK_PMD_HUGE(VADDR, REG1, REG2, FAIL_LABEL, 800f) \
sllx VADDR, 64 - PMD_SHIFT, REG2; \
srlx REG2, 64 - PAGE_SHIFT, REG2; \
- sllx REG1, PMD_PADDR_SHIFT, REG1; \
andn REG2, 0x7, REG2; \
add REG1, REG2, REG1; \
ldxa [REG1] ASI_PHYS_USE_EC, REG1; \