Fix 64bit SMP TLB handler and stack frame handling, optimize 32bit SMP

TLB handlers a bit, match definitions in pgtable-{32,64}.h better.

Signed-off-by: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

5 files changed, 34 insertions, 41 deletions

diff --git a/arch/mips/kernel/head.S b/arch/mips/kernel/head.S
index 124c27e908f4..9f6927169358 100644
--- a/arch/mips/kernel/head.S
+++ b/arch/mips/kernel/head.S
@@ -157,6 +157,7 @@ NESTED(kernel_entry, 16, sp)			# kernel entry point
         LONG_S          a2, fw_arg2
         LONG_S          a3, fw_arg3
 
+        MTC0            zero, CP0_CONTEXT       # clear context register
         PTR_LA          $28, init_thread_union
         PTR_ADDIU       sp, $28, _THREAD_SIZE - 32
         set_saved_sp    sp, t0, t1
diff --git a/arch/mips/mm/tlbex.c b/arch/mips/mm/tlbex.c
index bcf87a2b554c..018f7527a724 100644
--- a/arch/mips/mm/tlbex.c
+++ b/arch/mips/mm/tlbex.c
@@ -91,7 +91,7 @@ enum opcode {
         insn_addu, insn_addiu, insn_and, insn_andi, insn_beq,
         insn_beql, insn_bgez, insn_bgezl, insn_bltz, insn_bltzl,
         insn_bne, insn_daddu, insn_daddiu, insn_dmfc0, insn_dmtc0,
-        insn_dsll, insn_dsll32, insn_dsra, insn_dsrl, insn_dsrl32,
+        insn_dsll, insn_dsll32, insn_dsra, insn_dsrl,
         insn_dsubu, insn_eret, insn_j, insn_jal, insn_jr, insn_ld,
         insn_ll, insn_lld, insn_lui, insn_lw, insn_mfc0, insn_mtc0,
         insn_ori, insn_rfe, insn_sc, insn_scd, insn_sd, insn_sll,
@@ -134,7 +134,6 @@ static __initdata struct insn insn_table[] = {
         { insn_dsll32, M(spec_op,0,0,0,0,dsll32_op), RT | RD | RE },
         { insn_dsra, M(spec_op,0,0,0,0,dsra_op), RT | RD | RE },
         { insn_dsrl, M(spec_op,0,0,0,0,dsrl_op), RT | RD | RE },
-        { insn_dsrl32, M(spec_op,0,0,0,0,dsrl32_op), RT | RD | RE },
         { insn_dsubu, M(spec_op,0,0,0,0,dsubu_op), RS | RT | RD },
         { insn_eret, M(cop0_op,cop_op,0,0,0,eret_op), 0 },
         { insn_j, M(j_op,0,0,0,0,0), JIMM },
@@ -366,7 +365,6 @@ I_u2u1u3(_dsll);
 I_u2u1u3(_dsll32);
 I_u2u1u3(_dsra);
 I_u2u1u3(_dsrl);
-I_u2u1u3(_dsrl32);
 I_u3u1u2(_dsubu);
 I_0(_eret);
 I_u1(_j);
@@ -944,34 +942,29 @@ build_get_pmde64(u32 **p, struct label **l, struct reloc **r,
         /* No i_nop needed here, since the next insn doesn't touch TMP. */
 
 #ifdef CONFIG_SMP
+# ifdef CONFIG_BUILD_ELF64
         /*
-         * 64 bit SMP has the lower part of &pgd_current[smp_processor_id()]
+         * 64 bit SMP running in XKPHYS has smp_processor_id() << 3
          * stored in CONTEXT.
          */
-        if (in_compat_space_p(pgdc)) {
-                i_dmfc0(p, ptr, C0_CONTEXT);
-                i_dsra(p, ptr, ptr, 23);
-                i_ld(p, ptr, 0, ptr);
-        } else {
-#ifdef CONFIG_BUILD_ELF64
-                i_dmfc0(p, ptr, C0_CONTEXT);
-                i_dsrl(p, ptr, ptr, 23);
-                i_dsll(p, ptr, ptr, 3);
-                i_LA_mostly(p, tmp, pgdc);
-                i_daddu(p, ptr, ptr, tmp);
-                i_dmfc0(p, tmp, C0_BADVADDR);
-                i_ld(p, ptr, rel_lo(pgdc), ptr);
-#else
-                i_dmfc0(p, ptr, C0_CONTEXT);
-                i_lui(p, tmp, rel_highest(pgdc));
-                i_dsll(p, ptr, ptr, 9);
-                i_daddiu(p, tmp, tmp, rel_higher(pgdc));
-                i_dsrl32(p, ptr, ptr, 0);
-                i_and(p, ptr, ptr, tmp);
-                i_dmfc0(p, tmp, C0_BADVADDR);
-                i_ld(p, ptr, 0, ptr);
-#endif
-        }
+        i_dmfc0(p, ptr, C0_CONTEXT);
+        i_dsrl(p, ptr, ptr, 23);
+        i_LA_mostly(p, tmp, pgdc);
+        i_daddu(p, ptr, ptr, tmp);
+        i_dmfc0(p, tmp, C0_BADVADDR);
+        i_ld(p, ptr, rel_lo(pgdc), ptr);
+# else
+        /*
+         * 64 bit SMP running in compat space has the lower part of
+         * &pgd_current[smp_processor_id()] stored in CONTEXT.
+         */
+        if (!in_compat_space_p(pgdc))
+                panic("Invalid page directory address!");
+
+        i_dmfc0(p, ptr, C0_CONTEXT);
+        i_dsra(p, ptr, ptr, 23);
+        i_ld(p, ptr, 0, ptr);
+# endif
 #else
         i_LA_mostly(p, ptr, pgdc);
         i_ld(p, ptr, rel_lo(pgdc), ptr);
@@ -1028,7 +1021,6 @@ build_get_pgde32(u32 **p, unsigned int tmp, unsigned int ptr)
         i_mfc0(p, ptr, C0_CONTEXT);
         i_LA_mostly(p, tmp, pgdc);
         i_srl(p, ptr, ptr, 23);
-        i_sll(p, ptr, ptr, 2);
         i_addu(p, ptr, tmp, ptr);
 #else
         i_LA_mostly(p, ptr, pgdc);
diff --git a/include/asm-mips/mmu_context.h b/include/asm-mips/mmu_context.h
index 45cd72d172e8..19cdf7642e66 100644
--- a/include/asm-mips/mmu_context.h
+++ b/include/asm-mips/mmu_context.h
@@ -30,7 +30,7 @@ extern unsigned long pgd_current[];
 
 #ifdef CONFIG_32BIT
 #define TLBMISS_HANDLER_SETUP()                                         \
-        write_c0_context((unsigned long) smp_processor_id() << 23);     \
+        write_c0_context((unsigned long) smp_processor_id() << 25);     \
         TLBMISS_HANDLER_SETUP_PGD(swapper_pg_dir)
 #endif
 #if defined(CONFIG_64BIT) && !defined(CONFIG_BUILD_ELF64)
@@ -40,7 +40,7 @@ extern unsigned long pgd_current[];
 #endif
 #if defined(CONFIG_64BIT) && defined(CONFIG_BUILD_ELF64)
 #define TLBMISS_HANDLER_SETUP()                                         \
-        write_c0_context((unsigned long) smp_processor_id() << 23);     \
+        write_c0_context((unsigned long) smp_processor_id() << 26);     \
         TLBMISS_HANDLER_SETUP_PGD(swapper_pg_dir)
 #endif
 
diff --git a/include/asm-mips/pgtable-64.h b/include/asm-mips/pgtable-64.h
index 35007250ac2b..3e0a522c0f0e 100644
--- a/include/asm-mips/pgtable-64.h
+++ b/include/asm-mips/pgtable-64.h
@@ -114,7 +114,7 @@ extern pmd_t invalid_pmd_table[PTRS_PER_PMD];
 extern pmd_t empty_bad_pmd_table[PTRS_PER_PMD];
 
 /*
- * Empty pmd entries point to the invalid_pte_table.
+ * Empty pgd/pmd entries point to the invalid_pte_table.
  */
 static inline int pmd_none(pmd_t pmd)
 {
@@ -156,7 +156,8 @@ static inline void pud_clear(pud_t *pudp)
         pud_val(*pudp) = ((unsigned long) invalid_pmd_table);
 }
 
-#define pte_page(x)             pfn_to_page((unsigned long)((pte_val(x) >> PAGE_SHIFT)))
+#define pte_page(x)             pfn_to_page(pte_pfn(x))
+
 #ifdef CONFIG_CPU_VR41XX
 #define pte_pfn(x)              ((unsigned long)((x).pte >> (PAGE_SHIFT + 2)))
 #define pfn_pte(pfn, prot)      __pte(((pfn) << (PAGE_SHIFT + 2)) | pgprot_val(prot))
@@ -167,12 +168,14 @@ static inline void pud_clear(pud_t *pudp)
 
 #define __pgd_offset(address)   pgd_index(address)
 #define __pud_offset(address)   (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
+#define __pmd_offset(address)   pmd_index(address)
 #define page_pte(page) page_pte_prot(page, __pgprot(0))
 
 /* to find an entry in a kernel page-table-directory */
 #define pgd_offset_k(address) pgd_offset(&init_mm, 0)
 
 #define pgd_index(address)      (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
+#define pmd_index(address)      (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
 
 /* to find an entry in a page-table-directory */
 #define pgd_offset(mm,addr)     ((mm)->pgd + pgd_index(addr))
@@ -185,8 +188,7 @@ static inline unsigned long pud_page(pud_t pud)
 /* Find an entry in the second-level page table.. */
 static inline pmd_t *pmd_offset(pud_t * pud, unsigned long address)
 {
-        return (pmd_t *) pud_page(*pud) +
-               ((address >> PMD_SHIFT) & (PTRS_PER_PMD - 1));
+        return (pmd_t *) pud_page(*pud) + pmd_index(address);
 }
 
 /* Find an entry in the third-level page table.. */
diff --git a/include/asm-mips/stackframe.h b/include/asm-mips/stackframe.h
index e33e302c3429..de303e96260e 100644
--- a/include/asm-mips/stackframe.h
+++ b/include/asm-mips/stackframe.h
@@ -60,7 +60,6 @@
                 mfc0    k0, CP0_CONTEXT
                 lui     k1, %hi(kernelsp)
                 srl     k0, k0, 23
-                sll     k0, k0, 2
                 addu    k1, k0
                 LONG_L  k1, %lo(kernelsp)(k1)
 #endif
@@ -76,12 +75,12 @@
 #endif
 #if defined(CONFIG_64BIT) && defined(CONFIG_BUILD_ELF64)
                 MFC0    k1, CP0_CONTEXT
-                dsrl    k1, 23
-                dsll    k1, k1, 3
                 lui     k0, %highest(kernelsp)
+                dsrl    k1, 23
                 daddiu  k0, %higher(kernelsp)
                 dsll    k0, k0, 16
                 daddiu  k0, %hi(kernelsp)
+                dsll    k0, k0, 16
                 daddu   k1, k1, k0
                 LONG_L  k1, %lo(kernelsp)(k1)
 #endif
@@ -91,7 +90,6 @@
 #ifdef CONFIG_32BIT
                 mfc0    \temp, CP0_CONTEXT
                 srl     \temp, 23
-                sll     \temp, 2
                 LONG_S  \stackp, kernelsp(\temp)
 #endif
 #if defined(CONFIG_64BIT) && !defined(CONFIG_BUILD_ELF64)
@@ -102,8 +100,8 @@
                 LONG_S  \stackp, %lo(kernelsp)(\temp)
 #endif
 #if defined(CONFIG_64BIT) && defined(CONFIG_BUILD_ELF64)
-                lw      \temp, TI_CPU(gp)
-                dsll    \temp, 3
+                MFC0    \temp, CP0_CONTEXT
+                dsrl    \temp, 23
                 LONG_S  \stackp, kernelsp(\temp)
 #endif
                 .endm