Merge branch 'master' into next

7 files changed, 135 insertions, 44 deletions

diff --git a/arch/mips/mm/cache.c b/arch/mips/mm/cache.c
index be8627bc5b02..12af739048fa 100644
--- a/arch/mips/mm/cache.c
+++ b/arch/mips/mm/cache.c
@@ -133,7 +133,7 @@ void __update_cache(struct vm_area_struct *vma, unsigned long address,
 }
 
 unsigned long _page_cachable_default;
-EXPORT_SYMBOL_GPL(_page_cachable_default);
+EXPORT_SYMBOL(_page_cachable_default);
 
 static inline void setup_protection_map(void)
 {
diff --git a/arch/mips/mm/dma-default.c b/arch/mips/mm/dma-default.c
index 9367e33fbd18..9547bc0cf188 100644
--- a/arch/mips/mm/dma-default.c
+++ b/arch/mips/mm/dma-default.c
@@ -14,6 +14,7 @@
 #include <linux/module.h>
 #include <linux/scatterlist.h>
 #include <linux/string.h>
+#include <linux/gfp.h>
 
 #include <asm/cache.h>
 #include <asm/io.h>
diff --git a/arch/mips/mm/hugetlbpage.c b/arch/mips/mm/hugetlbpage.c
index cd0660c51f28..a7fee0dfb7a9 100644
--- a/arch/mips/mm/hugetlbpage.c
+++ b/arch/mips/mm/hugetlbpage.c
@@ -16,7 +16,6 @@
 #include <linux/mm.h>
 #include <linux/hugetlb.h>
 #include <linux/pagemap.h>
-#include <linux/slab.h>
 #include <linux/err.h>
 #include <linux/sysctl.h>
 #include <asm/mman.h>
diff --git a/arch/mips/mm/init.c b/arch/mips/mm/init.c
index 12539af38a99..2efcbd24c82f 100644
--- a/arch/mips/mm/init.c
+++ b/arch/mips/mm/init.c
@@ -28,6 +28,7 @@
 #include <linux/proc_fs.h>
 #include <linux/pfn.h>
 #include <linux/hardirq.h>
+#include <linux/gfp.h>
 
 #include <asm/asm-offsets.h>
 #include <asm/bootinfo.h>
diff --git a/arch/mips/mm/ioremap.c b/arch/mips/mm/ioremap.c
index 0c43248347bd..cacfd31e8ec9 100644
--- a/arch/mips/mm/ioremap.c
+++ b/arch/mips/mm/ioremap.c
@@ -10,6 +10,7 @@
 #include <asm/addrspace.h>
 #include <asm/byteorder.h>
 #include <linux/sched.h>
+#include <linux/slab.h>
 #include <linux/vmalloc.h>
 #include <asm/cacheflush.h>
 #include <asm/io.h>
diff --git a/arch/mips/mm/tlbex.c b/arch/mips/mm/tlbex.c
index 0de0e4127d66..86f004dc8355 100644
--- a/arch/mips/mm/tlbex.c
+++ b/arch/mips/mm/tlbex.c
@@ -31,6 +31,16 @@
 #include <asm/war.h>
 #include <asm/uasm.h>
 
+/*
+ * TLB load/store/modify handlers.
+ *
+ * Only the fastpath gets synthesized at runtime, the slowpath for
+ * do_page_fault remains normal asm.
+ */
+extern void tlb_do_page_fault_0(void);
+extern void tlb_do_page_fault_1(void);
+
+
 static inline int r45k_bvahwbug(void)
 {
 	/* XXX: We should probe for the presence of this bug, but we don't. */
@@ -83,6 +93,7 @@ enum label_id {
 	label_nopage_tlbm,
 	label_smp_pgtable_change,
 	label_r3000_write_probe_fail,
+	label_large_segbits_fault,
 #ifdef CONFIG_HUGETLB_PAGE
 	label_tlb_huge_update,
 #endif
@@ -101,6 +112,7 @@ UASM_L_LA(_nopage_tlbs)
 UASM_L_LA(_nopage_tlbm)
 UASM_L_LA(_smp_pgtable_change)
 UASM_L_LA(_r3000_write_probe_fail)
+UASM_L_LA(_large_segbits_fault)
 #ifdef CONFIG_HUGETLB_PAGE
 UASM_L_LA(_tlb_huge_update)
 #endif
@@ -157,6 +169,10 @@ static u32 tlb_handler[128] __cpuinitdata;
 static struct uasm_label labels[128] __cpuinitdata;
 static struct uasm_reloc relocs[128] __cpuinitdata;
 
+#ifdef CONFIG_64BIT
+static int check_for_high_segbits __cpuinitdata;
+#endif
+
 #ifndef CONFIG_MIPS_PGD_C0_CONTEXT
 /*
  * CONFIG_MIPS_PGD_C0_CONTEXT implies 64 bit and lack of pgd_current,
@@ -408,7 +424,7 @@ static __cpuinit __maybe_unused void build_convert_pte_to_entrylo(u32 **p,
 		UASM_i_ROTR(p, reg, reg, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC));
 	} else {
 #ifdef CONFIG_64BIT_PHYS_ADDR
-		uasm_i_dsrl(p, reg, reg, ilog2(_PAGE_GLOBAL));
+		uasm_i_dsrl_safe(p, reg, reg, ilog2(_PAGE_GLOBAL));
 #else
 		UASM_i_SRL(p, reg, reg, ilog2(_PAGE_GLOBAL));
 #endif
@@ -532,7 +548,24 @@ build_get_pmde64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
 	 * The vmalloc handling is not in the hotpath.
 	 */
 	uasm_i_dmfc0(p, tmp, C0_BADVADDR);
-	uasm_il_bltz(p, r, tmp, label_vmalloc);
+
+	if (check_for_high_segbits) {
+		/*
+		 * The kernel currently implicitely assumes that the
+		 * MIPS SEGBITS parameter for the processor is
+		 * (PGDIR_SHIFT+PGDIR_BITS) or less, and will never
+		 * allocate virtual addresses outside the maximum
+		 * range for SEGBITS = (PGDIR_SHIFT+PGDIR_BITS). But
+		 * that doesn't prevent user code from accessing the
+		 * higher xuseg addresses.  Here, we make sure that
+		 * everything but the lower xuseg addresses goes down
+		 * the module_alloc/vmalloc path.
+		 */
+		uasm_i_dsrl_safe(p, ptr, tmp, PGDIR_SHIFT + PGD_ORDER + PAGE_SHIFT - 3);
+		uasm_il_bnez(p, r, ptr, label_vmalloc);
+	} else {
+		uasm_il_bltz(p, r, tmp, label_vmalloc);
+	}
 	/* No uasm_i_nop needed here, since the next insn doesn't touch TMP. */
 
 #ifdef CONFIG_MIPS_PGD_C0_CONTEXT
@@ -549,14 +582,14 @@ build_get_pmde64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
 	 * SMTC uses TCBind value as "CPU" index
 	 */
 	uasm_i_mfc0(p, ptr, C0_TCBIND);
-	uasm_i_dsrl(p, ptr, ptr, 19);
+	uasm_i_dsrl_safe(p, ptr, ptr, 19);
 # else
 	/*
 	 * 64 bit SMP running in XKPHYS has smp_processor_id() << 3
 	 * stored in CONTEXT.
 	 */
 	uasm_i_dmfc0(p, ptr, C0_CONTEXT);
-	uasm_i_dsrl(p, ptr, ptr, 23);
+	uasm_i_dsrl_safe(p, ptr, ptr, 23);
 # endif
 	UASM_i_LA_mostly(p, tmp, pgdc);
 	uasm_i_daddu(p, ptr, ptr, tmp);
@@ -569,44 +602,78 @@ build_get_pmde64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
 
 	uasm_l_vmalloc_done(l, *p);
 
-	if (PGDIR_SHIFT - 3 < 32)		/* get pgd offset in bytes */
-		uasm_i_dsrl(p, tmp, tmp, PGDIR_SHIFT-3);
-	else
-		uasm_i_dsrl32(p, tmp, tmp, PGDIR_SHIFT - 3 - 32);
+	/* get pgd offset in bytes */
+	uasm_i_dsrl_safe(p, tmp, tmp, PGDIR_SHIFT - 3);
 
 	uasm_i_andi(p, tmp, tmp, (PTRS_PER_PGD - 1)<<3);
 	uasm_i_daddu(p, ptr, ptr, tmp); /* add in pgd offset */
 #ifndef __PAGETABLE_PMD_FOLDED
 	uasm_i_dmfc0(p, tmp, C0_BADVADDR); /* get faulting address */
 	uasm_i_ld(p, ptr, 0, ptr); /* get pmd pointer */
-	uasm_i_dsrl(p, tmp, tmp, PMD_SHIFT-3); /* get pmd offset in bytes */
+	uasm_i_dsrl_safe(p, tmp, tmp, PMD_SHIFT-3); /* get pmd offset in bytes */
 	uasm_i_andi(p, tmp, tmp, (PTRS_PER_PMD - 1)<<3);
 	uasm_i_daddu(p, ptr, ptr, tmp); /* add in pmd offset */
 #endif
 }
 
+enum vmalloc64_mode {not_refill, refill};
 /*
  * BVADDR is the faulting address, PTR is scratch.
  * PTR will hold the pgd for vmalloc.
  */
 static void __cpuinit
 build_get_pgd_vmalloc64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
-			unsigned int bvaddr, unsigned int ptr)
+			unsigned int bvaddr, unsigned int ptr,
+			enum vmalloc64_mode mode)
 {
 	long swpd = (long)swapper_pg_dir;
+	int single_insn_swpd;
+	int did_vmalloc_branch = 0;
+
+	single_insn_swpd = uasm_in_compat_space_p(swpd) && !uasm_rel_lo(swpd);
 
 	uasm_l_vmalloc(l, *p);
 
-	if (uasm_in_compat_space_p(swpd) && !uasm_rel_lo(swpd)) {
-		uasm_il_b(p, r, label_vmalloc_done);
-		uasm_i_lui(p, ptr, uasm_rel_hi(swpd));
-	} else {
-		UASM_i_LA_mostly(p, ptr, swpd);
-		uasm_il_b(p, r, label_vmalloc_done);
-		if (uasm_in_compat_space_p(swpd))
-			uasm_i_addiu(p, ptr, ptr, uasm_rel_lo(swpd));
-		else
-			uasm_i_daddiu(p, ptr, ptr, uasm_rel_lo(swpd));
+	if (mode == refill && check_for_high_segbits) {
+		if (single_insn_swpd) {
+			uasm_il_bltz(p, r, bvaddr, label_vmalloc_done);
+			uasm_i_lui(p, ptr, uasm_rel_hi(swpd));
+			did_vmalloc_branch = 1;
+			/* fall through */
+		} else {
+			uasm_il_bgez(p, r, bvaddr, label_large_segbits_fault);
+		}
+	}
+	if (!did_vmalloc_branch) {
+		if (uasm_in_compat_space_p(swpd) && !uasm_rel_lo(swpd)) {
+			uasm_il_b(p, r, label_vmalloc_done);
+			uasm_i_lui(p, ptr, uasm_rel_hi(swpd));
+		} else {
+			UASM_i_LA_mostly(p, ptr, swpd);
+			uasm_il_b(p, r, label_vmalloc_done);
+			if (uasm_in_compat_space_p(swpd))
+				uasm_i_addiu(p, ptr, ptr, uasm_rel_lo(swpd));
+			else
+				uasm_i_daddiu(p, ptr, ptr, uasm_rel_lo(swpd));
+		}
+	}
+	if (mode == refill && check_for_high_segbits) {
+		uasm_l_large_segbits_fault(l, *p);
+		/*
+		 * We get here if we are an xsseg address, or if we are
+		 * an xuseg address above (PGDIR_SHIFT+PGDIR_BITS) boundary.
+		 *
+		 * Ignoring xsseg (assume disabled so would generate
+		 * (address errors?), the only remaining possibility
+		 * is the upper xuseg addresses.  On processors with
+		 * TLB_SEGBITS <= PGDIR_SHIFT+PGDIR_BITS, these
+		 * addresses would have taken an address error. We try
+		 * to mimic that here by taking a load/istream page
+		 * fault.
+		 */
+		UASM_i_LA(p, ptr, (unsigned long)tlb_do_page_fault_0);
+		uasm_i_jr(p, ptr);
+		uasm_i_nop(p);
 	}
 }
 
@@ -720,9 +787,9 @@ static void __cpuinit build_update_entries(u32 **p, unsigned int tmp,
 			UASM_i_MTC0(p, tmp, C0_ENTRYLO0); /* load it */
 			UASM_i_ROTR(p, ptep, ptep, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC));
 		} else {
-			uasm_i_dsrl(p, tmp, tmp, ilog2(_PAGE_GLOBAL)); /* convert to entrylo0 */
+			uasm_i_dsrl_safe(p, tmp, tmp, ilog2(_PAGE_GLOBAL)); /* convert to entrylo0 */
 			UASM_i_MTC0(p, tmp, C0_ENTRYLO0); /* load it */
-			uasm_i_dsrl(p, ptep, ptep, ilog2(_PAGE_GLOBAL)); /* convert to entrylo1 */
+			uasm_i_dsrl_safe(p, ptep, ptep, ilog2(_PAGE_GLOBAL)); /* convert to entrylo1 */
 		}
 		UASM_i_MTC0(p, ptep, C0_ENTRYLO1); /* load it */
 	} else {
@@ -788,10 +855,15 @@ static void __cpuinit build_r4000_tlb_refill_handler(void)
 	 * create the plain linear handler
 	 */
 	if (bcm1250_m3_war()) {
-		UASM_i_MFC0(&p, K0, C0_BADVADDR);
-		UASM_i_MFC0(&p, K1, C0_ENTRYHI);
+		unsigned int segbits = 44;
+
+		uasm_i_dmfc0(&p, K0, C0_BADVADDR);
+		uasm_i_dmfc0(&p, K1, C0_ENTRYHI);
 		uasm_i_xor(&p, K0, K0, K1);
-		UASM_i_SRL(&p, K0, K0, PAGE_SHIFT + 1);
+		uasm_i_dsrl_safe(&p, K1, K0, 62);
+		uasm_i_dsrl_safe(&p, K0, K0, 12 + 1);
+		uasm_i_dsll_safe(&p, K0, K0, 64 + 12 + 1 - segbits);
+		uasm_i_or(&p, K0, K0, K1);
 		uasm_il_bnez(&p, &r, K0, label_leave);
 		/* No need for uasm_i_nop */
 	}
@@ -820,7 +892,7 @@ static void __cpuinit build_r4000_tlb_refill_handler(void)
 #endif
 
 #ifdef CONFIG_64BIT
-	build_get_pgd_vmalloc64(&p, &l, &r, K0, K1);
+	build_get_pgd_vmalloc64(&p, &l, &r, K0, K1, refill);
 #endif
 
 	/*
@@ -930,15 +1002,6 @@ static void __cpuinit build_r4000_tlb_refill_handler(void)
 }
 
 /*
- * TLB load/store/modify handlers.
- *
- * Only the fastpath gets synthesized at runtime, the slowpath for
- * do_page_fault remains normal asm.
- */
-extern void tlb_do_page_fault_0(void);
-extern void tlb_do_page_fault_1(void);
-
-/*
  * 128 instructions for the fastpath handler is generous and should
  * never be exceeded.
  */
@@ -1297,7 +1360,7 @@ build_r4000_tlbchange_handler_tail(u32 **p, struct uasm_label **l,
 	uasm_i_eret(p); /* return from trap */
 
 #ifdef CONFIG_64BIT
-	build_get_pgd_vmalloc64(p, l, r, tmp, ptr);
+	build_get_pgd_vmalloc64(p, l, r, tmp, ptr, not_refill);
 #endif
 }
 
@@ -1312,10 +1375,15 @@ static void __cpuinit build_r4000_tlb_load_handler(void)
 	memset(relocs, 0, sizeof(relocs));
 
 	if (bcm1250_m3_war()) {
-		UASM_i_MFC0(&p, K0, C0_BADVADDR);
-		UASM_i_MFC0(&p, K1, C0_ENTRYHI);
+		unsigned int segbits = 44;
+
+		uasm_i_dmfc0(&p, K0, C0_BADVADDR);
+		uasm_i_dmfc0(&p, K1, C0_ENTRYHI);
 		uasm_i_xor(&p, K0, K0, K1);
-		UASM_i_SRL(&p, K0, K0, PAGE_SHIFT + 1);
+		uasm_i_dsrl_safe(&p, K1, K0, 62);
+		uasm_i_dsrl_safe(&p, K0, K0, 12 + 1);
+		uasm_i_dsll_safe(&p, K0, K0, 64 + 12 + 1 - segbits);
+		uasm_i_or(&p, K0, K0, K1);
 		uasm_il_bnez(&p, &r, K0, label_leave);
 		/* No need for uasm_i_nop */
 	}
@@ -1516,6 +1584,10 @@ void __cpuinit build_tlb_refill_handler(void)
 	 */
 	static int run_once = 0;
 
+#ifdef CONFIG_64BIT
+	check_for_high_segbits = current_cpu_data.vmbits > (PGDIR_SHIFT + PGD_ORDER + PAGE_SHIFT - 3);
+#endif
+
 	switch (current_cpu_type()) {
 	case CPU_R2000:
 	case CPU_R3000:
diff --git a/arch/mips/mm/uasm.c b/arch/mips/mm/uasm.c
index 1581e9852461..611d564fdcf1 100644
--- a/arch/mips/mm/uasm.c
+++ b/arch/mips/mm/uasm.c
@@ -31,7 +31,8 @@ enum fields {
 	BIMM = 0x040,
 	JIMM = 0x080,
 	FUNC = 0x100,
-	SET = 0x200
+	SET = 0x200,
+	SCIMM = 0x400
 };
 
 #define OP_MASK		0x3f
@@ -52,6 +53,8 @@ enum fields {
 #define FUNC_SH		0
 #define SET_MASK	0x7
 #define SET_SH		0
+#define SCIMM_MASK	0xfffff
+#define SCIMM_SH	6
 
 enum opcode {
 	insn_invalid,
@@ -61,10 +64,10 @@ enum opcode {
 	insn_dmtc0, insn_dsll, insn_dsll32, insn_dsra, insn_dsrl,
 	insn_dsrl32, insn_drotr, 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_pref, insn_rfe, insn_sc, insn_scd,
+	insn_mtc0, insn_or, insn_ori, insn_pref, insn_rfe, insn_sc, insn_scd,
 	insn_sd, insn_sll, insn_sra, insn_srl, insn_rotr, insn_subu, insn_sw,
 	insn_tlbp, insn_tlbr, insn_tlbwi, insn_tlbwr, insn_xor, insn_xori,
-	insn_dins
+	insn_dins, insn_syscall
 };
 
 struct insn {
@@ -117,6 +120,7 @@ static struct insn insn_table[] __cpuinitdata = {
 	{ insn_lw,  M(lw_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
 	{ insn_mfc0,  M(cop0_op, mfc_op, 0, 0, 0, 0),  RT | RD | SET},
 	{ insn_mtc0,  M(cop0_op, mtc_op, 0, 0, 0, 0),  RT | RD | SET},
+	{ insn_or,  M(spec_op, 0, 0, 0, 0, or_op),  RS | RT | RD },
 	{ insn_ori,  M(ori_op, 0, 0, 0, 0, 0),  RS | RT | UIMM },
 	{ insn_pref,  M(pref_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
 	{ insn_rfe,  M(cop0_op, cop_op, 0, 0, 0, rfe_op),  0 },
@@ -136,6 +140,7 @@ static struct insn insn_table[] __cpuinitdata = {
 	{ insn_xor,  M(spec_op, 0, 0, 0, 0, xor_op),  RS | RT | RD },
 	{ insn_xori,  M(xori_op, 0, 0, 0, 0, 0),  RS | RT | UIMM },
 	{ insn_dins, M(spec3_op, 0, 0, 0, 0, dins_op), RS | RT | RD | RE },
+	{ insn_syscall, M(spec_op, 0, 0, 0, 0, syscall_op), SCIMM},
 	{ insn_invalid, 0, 0 }
 };
 
@@ -208,6 +213,14 @@ static inline __cpuinit u32 build_jimm(u32 arg)
 	return (arg >> 2) & JIMM_MASK;
 }
 
+static inline __cpuinit u32 build_scimm(u32 arg)
+{
+	if (arg & ~SCIMM_MASK)
+		printk(KERN_WARNING "Micro-assembler field overflow\n");
+
+	return (arg & SCIMM_MASK) << SCIMM_SH;
+}
+
 static inline __cpuinit u32 build_func(u32 arg)
 {
 	if (arg & ~FUNC_MASK)
@@ -266,6 +279,8 @@ static void __cpuinit build_insn(u32 **buf, enum opcode opc, ...)
 		op |= build_func(va_arg(ap, u32));
 	if (ip->fields & SET)
 		op |= build_set(va_arg(ap, u32));
+	if (ip->fields & SCIMM)
+		op |= build_scimm(va_arg(ap, u32));
 	va_end(ap);
 
 	**buf = op;
@@ -373,6 +388,7 @@ I_u2s3u1(_lw)
 I_u1u2u3(_mfc0)
 I_u1u2u3(_mtc0)
 I_u2u1u3(_ori)
+I_u3u1u2(_or)
 I_u2s3u1(_pref)
 I_0(_rfe)
 I_u2s3u1(_sc)
@@ -391,6 +407,7 @@ I_0(_tlbwr)
 I_u3u1u2(_xor)
 I_u2u1u3(_xori)
 I_u2u1msbu3(_dins);
+I_u1(_syscall);
 
 /* Handle labels. */
 void __cpuinit uasm_build_label(struct uasm_label **lab, u32 *addr, int lid)