From 1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 Mon Sep 17 00:00:00 2001 From: Linus Torvalds Date: Sat, 16 Apr 2005 15:20:36 -0700 Subject: Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip! --- arch/ppc64/mm/Makefile | 11 + arch/ppc64/mm/fault.c | 312 +++++++++++++++ arch/ppc64/mm/hash_low.S | 287 ++++++++++++++ arch/ppc64/mm/hash_native.c | 423 ++++++++++++++++++++ arch/ppc64/mm/hash_utils.c | 439 +++++++++++++++++++++ arch/ppc64/mm/hugetlbpage.c | 904 ++++++++++++++++++++++++++++++++++++++++++ arch/ppc64/mm/imalloc.c | 312 +++++++++++++++ arch/ppc64/mm/init.c | 927 ++++++++++++++++++++++++++++++++++++++++++++ arch/ppc64/mm/mmap.c | 86 ++++ arch/ppc64/mm/numa.c | 734 +++++++++++++++++++++++++++++++++++ arch/ppc64/mm/slb.c | 159 ++++++++ arch/ppc64/mm/slb_low.S | 154 ++++++++ arch/ppc64/mm/stab.c | 239 ++++++++++++ arch/ppc64/mm/tlb.c | 180 +++++++++ 14 files changed, 5167 insertions(+) create mode 100644 arch/ppc64/mm/Makefile create mode 100644 arch/ppc64/mm/fault.c create mode 100644 arch/ppc64/mm/hash_low.S create mode 100644 arch/ppc64/mm/hash_native.c create mode 100644 arch/ppc64/mm/hash_utils.c create mode 100644 arch/ppc64/mm/hugetlbpage.c create mode 100644 arch/ppc64/mm/imalloc.c create mode 100644 arch/ppc64/mm/init.c create mode 100644 arch/ppc64/mm/mmap.c create mode 100644 arch/ppc64/mm/numa.c create mode 100644 arch/ppc64/mm/slb.c create mode 100644 arch/ppc64/mm/slb_low.S create mode 100644 arch/ppc64/mm/stab.c create mode 100644 arch/ppc64/mm/tlb.c (limited to 'arch/ppc64/mm') diff --git a/arch/ppc64/mm/Makefile b/arch/ppc64/mm/Makefile new file mode 100644 index 00000000000..ac522d57b2a --- /dev/null +++ b/arch/ppc64/mm/Makefile @@ -0,0 +1,11 @@ +# +# Makefile for the linux ppc-specific parts of the memory manager. +# + +EXTRA_CFLAGS += -mno-minimal-toc + +obj-y := fault.o init.o imalloc.o hash_utils.o hash_low.o tlb.o \ + slb_low.o slb.o stab.o mmap.o +obj-$(CONFIG_DISCONTIGMEM) += numa.o +obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o +obj-$(CONFIG_PPC_MULTIPLATFORM) += hash_native.o diff --git a/arch/ppc64/mm/fault.c b/arch/ppc64/mm/fault.c new file mode 100644 index 00000000000..20b0f37e8bf --- /dev/null +++ b/arch/ppc64/mm/fault.c @@ -0,0 +1,312 @@ +/* + * arch/ppc/mm/fault.c + * + * PowerPC version + * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org) + * + * Derived from "arch/i386/mm/fault.c" + * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds + * + * Modified by Cort Dougan and Paul Mackerras. + * + * Modified for PPC64 by Dave Engebretsen (engebret@ibm.com) + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include +#include + +/* + * Check whether the instruction at regs->nip is a store using + * an update addressing form which will update r1. + */ +static int store_updates_sp(struct pt_regs *regs) +{ + unsigned int inst; + + if (get_user(inst, (unsigned int __user *)regs->nip)) + return 0; + /* check for 1 in the rA field */ + if (((inst >> 16) & 0x1f) != 1) + return 0; + /* check major opcode */ + switch (inst >> 26) { + case 37: /* stwu */ + case 39: /* stbu */ + case 45: /* sthu */ + case 53: /* stfsu */ + case 55: /* stfdu */ + return 1; + case 62: /* std or stdu */ + return (inst & 3) == 1; + case 31: + /* check minor opcode */ + switch ((inst >> 1) & 0x3ff) { + case 181: /* stdux */ + case 183: /* stwux */ + case 247: /* stbux */ + case 439: /* sthux */ + case 695: /* stfsux */ + case 759: /* stfdux */ + return 1; + } + } + return 0; +} + +/* + * The error_code parameter is + * - DSISR for a non-SLB data access fault, + * - SRR1 & 0x08000000 for a non-SLB instruction access fault + * - 0 any SLB fault. + * The return value is 0 if the fault was handled, or the signal + * number if this is a kernel fault that can't be handled here. + */ +int do_page_fault(struct pt_regs *regs, unsigned long address, + unsigned long error_code) +{ + struct vm_area_struct * vma; + struct mm_struct *mm = current->mm; + siginfo_t info; + unsigned long code = SEGV_MAPERR; + unsigned long is_write = error_code & DSISR_ISSTORE; + unsigned long trap = TRAP(regs); + unsigned long is_exec = trap == 0x400; + + BUG_ON((trap == 0x380) || (trap == 0x480)); + + if (notify_die(DIE_PAGE_FAULT, "page_fault", regs, error_code, + 11, SIGSEGV) == NOTIFY_STOP) + return 0; + + if (trap == 0x300) { + if (debugger_fault_handler(regs)) + return 0; + } + + /* On a kernel SLB miss we can only check for a valid exception entry */ + if (!user_mode(regs) && (address >= TASK_SIZE)) + return SIGSEGV; + + if (error_code & DSISR_DABRMATCH) { + if (notify_die(DIE_DABR_MATCH, "dabr_match", regs, error_code, + 11, SIGSEGV) == NOTIFY_STOP) + return 0; + if (debugger_dabr_match(regs)) + return 0; + } + + if (in_atomic() || mm == NULL) { + if (!user_mode(regs)) + return SIGSEGV; + /* in_atomic() in user mode is really bad, + as is current->mm == NULL. */ + printk(KERN_EMERG "Page fault in user mode with" + "in_atomic() = %d mm = %p\n", in_atomic(), mm); + printk(KERN_EMERG "NIP = %lx MSR = %lx\n", + regs->nip, regs->msr); + die("Weird page fault", regs, SIGSEGV); + } + + /* When running in the kernel we expect faults to occur only to + * addresses in user space. All other faults represent errors in the + * kernel and should generate an OOPS. Unfortunatly, in the case of an + * erroneous fault occuring in a code path which already holds mmap_sem + * we will deadlock attempting to validate the fault against the + * address space. Luckily the kernel only validly references user + * space from well defined areas of code, which are listed in the + * exceptions table. + * + * As the vast majority of faults will be valid we will only perform + * the source reference check when there is a possibilty of a deadlock. + * Attempt to lock the address space, if we cannot we then validate the + * source. If this is invalid we can skip the address space check, + * thus avoiding the deadlock. + */ + if (!down_read_trylock(&mm->mmap_sem)) { + if (!user_mode(regs) && !search_exception_tables(regs->nip)) + goto bad_area_nosemaphore; + + down_read(&mm->mmap_sem); + } + + vma = find_vma(mm, address); + if (!vma) + goto bad_area; + + if (vma->vm_start <= address) { + goto good_area; + } + if (!(vma->vm_flags & VM_GROWSDOWN)) + goto bad_area; + + /* + * N.B. The POWER/Open ABI allows programs to access up to + * 288 bytes below the stack pointer. + * The kernel signal delivery code writes up to about 1.5kB + * below the stack pointer (r1) before decrementing it. + * The exec code can write slightly over 640kB to the stack + * before setting the user r1. Thus we allow the stack to + * expand to 1MB without further checks. + */ + if (address + 0x100000 < vma->vm_end) { + /* get user regs even if this fault is in kernel mode */ + struct pt_regs *uregs = current->thread.regs; + if (uregs == NULL) + goto bad_area; + + /* + * A user-mode access to an address a long way below + * the stack pointer is only valid if the instruction + * is one which would update the stack pointer to the + * address accessed if the instruction completed, + * i.e. either stwu rs,n(r1) or stwux rs,r1,rb + * (or the byte, halfword, float or double forms). + * + * If we don't check this then any write to the area + * between the last mapped region and the stack will + * expand the stack rather than segfaulting. + */ + if (address + 2048 < uregs->gpr[1] + && (!user_mode(regs) || !store_updates_sp(regs))) + goto bad_area; + } + + if (expand_stack(vma, address)) + goto bad_area; + +good_area: + code = SEGV_ACCERR; + + if (is_exec) { + /* protection fault */ + if (error_code & DSISR_PROTFAULT) + goto bad_area; + if (!(vma->vm_flags & VM_EXEC)) + goto bad_area; + /* a write */ + } else if (is_write) { + if (!(vma->vm_flags & VM_WRITE)) + goto bad_area; + /* a read */ + } else { + if (!(vma->vm_flags & VM_READ)) + goto bad_area; + } + + survive: + /* + * If for any reason at all we couldn't handle the fault, + * make sure we exit gracefully rather than endlessly redo + * the fault. + */ + switch (handle_mm_fault(mm, vma, address, is_write)) { + + case VM_FAULT_MINOR: + current->min_flt++; + break; + case VM_FAULT_MAJOR: + current->maj_flt++; + break; + case VM_FAULT_SIGBUS: + goto do_sigbus; + case VM_FAULT_OOM: + goto out_of_memory; + default: + BUG(); + } + + up_read(&mm->mmap_sem); + return 0; + +bad_area: + up_read(&mm->mmap_sem); + +bad_area_nosemaphore: + /* User mode accesses cause a SIGSEGV */ + if (user_mode(regs)) { + info.si_signo = SIGSEGV; + info.si_errno = 0; + info.si_code = code; + info.si_addr = (void __user *) address; + force_sig_info(SIGSEGV, &info, current); + return 0; + } + + if (trap == 0x400 && (error_code & DSISR_PROTFAULT) + && printk_ratelimit()) + printk(KERN_CRIT "kernel tried to execute NX-protected" + " page (%lx) - exploit attempt? (uid: %d)\n", + address, current->uid); + + return SIGSEGV; + +/* + * We ran out of memory, or some other thing happened to us that made + * us unable to handle the page fault gracefully. + */ +out_of_memory: + up_read(&mm->mmap_sem); + if (current->pid == 1) { + yield(); + down_read(&mm->mmap_sem); + goto survive; + } + printk("VM: killing process %s\n", current->comm); + if (user_mode(regs)) + do_exit(SIGKILL); + return SIGKILL; + +do_sigbus: + up_read(&mm->mmap_sem); + if (user_mode(regs)) { + info.si_signo = SIGBUS; + info.si_errno = 0; + info.si_code = BUS_ADRERR; + info.si_addr = (void __user *)address; + force_sig_info(SIGBUS, &info, current); + return 0; + } + return SIGBUS; +} + +/* + * bad_page_fault is called when we have a bad access from the kernel. + * It is called from do_page_fault above and from some of the procedures + * in traps.c. + */ +void bad_page_fault(struct pt_regs *regs, unsigned long address, int sig) +{ + const struct exception_table_entry *entry; + + /* Are we prepared to handle this fault? */ + if ((entry = search_exception_tables(regs->nip)) != NULL) { + regs->nip = entry->fixup; + return; + } + + /* kernel has accessed a bad area */ + die("Kernel access of bad area", regs, sig); +} diff --git a/arch/ppc64/mm/hash_low.S b/arch/ppc64/mm/hash_low.S new file mode 100644 index 00000000000..8c0156a3700 --- /dev/null +++ b/arch/ppc64/mm/hash_low.S @@ -0,0 +1,287 @@ +/* + * ppc64 MMU hashtable management routines + * + * (c) Copyright IBM Corp. 2003 + * + * Maintained by: Benjamin Herrenschmidt + * + * + * This file is covered by the GNU Public Licence v2 as + * described in the kernel's COPYING file. + */ + +#include +#include +#include +#include +#include +#include +#include +#include + + .text + +/* + * Stackframe: + * + * +-> Back chain (SP + 256) + * | General register save area (SP + 112) + * | Parameter save area (SP + 48) + * | TOC save area (SP + 40) + * | link editor doubleword (SP + 32) + * | compiler doubleword (SP + 24) + * | LR save area (SP + 16) + * | CR save area (SP + 8) + * SP ---> +-- Back chain (SP + 0) + */ +#define STACKFRAMESIZE 256 + +/* Save parameters offsets */ +#define STK_PARM(i) (STACKFRAMESIZE + 48 + ((i)-3)*8) + +/* Save non-volatile offsets */ +#define STK_REG(i) (112 + ((i)-14)*8) + +/* + * _hash_page(unsigned long ea, unsigned long access, unsigned long vsid, + * pte_t *ptep, unsigned long trap, int local) + * + * Adds a page to the hash table. This is the non-LPAR version for now + */ + +_GLOBAL(__hash_page) + mflr r0 + std r0,16(r1) + stdu r1,-STACKFRAMESIZE(r1) + /* Save all params that we need after a function call */ + std r6,STK_PARM(r6)(r1) + std r8,STK_PARM(r8)(r1) + + /* Add _PAGE_PRESENT to access */ + ori r4,r4,_PAGE_PRESENT + + /* Save non-volatile registers. + * r31 will hold "old PTE" + * r30 is "new PTE" + * r29 is "va" + * r28 is a hash value + * r27 is hashtab mask (maybe dynamic patched instead ?) + */ + std r27,STK_REG(r27)(r1) + std r28,STK_REG(r28)(r1) + std r29,STK_REG(r29)(r1) + std r30,STK_REG(r30)(r1) + std r31,STK_REG(r31)(r1) + + /* Step 1: + * + * Check permissions, atomically mark the linux PTE busy + * and hashed. + */ +1: + ldarx r31,0,r6 + /* Check access rights (access & ~(pte_val(*ptep))) */ + andc. r0,r4,r31 + bne- htab_wrong_access + /* Check if PTE is busy */ + andi. r0,r31,_PAGE_BUSY + bne- 1b + /* Prepare new PTE value (turn access RW into DIRTY, then + * add BUSY,HASHPTE and ACCESSED) + */ + rlwinm r30,r4,32-9+7,31-7,31-7 /* _PAGE_RW -> _PAGE_DIRTY */ + or r30,r30,r31 + ori r30,r30,_PAGE_BUSY | _PAGE_ACCESSED | _PAGE_HASHPTE + /* Write the linux PTE atomically (setting busy) */ + stdcx. r30,0,r6 + bne- 1b + isync + + /* Step 2: + * + * Insert/Update the HPTE in the hash table. At this point, + * r4 (access) is re-useable, we use it for the new HPTE flags + */ + + /* Calc va and put it in r29 */ + rldicr r29,r5,28,63-28 + rldicl r3,r3,0,36 + or r29,r3,r29 + + /* Calculate hash value for primary slot and store it in r28 */ + rldicl r5,r5,0,25 /* vsid & 0x0000007fffffffff */ + rldicl r0,r3,64-12,48 /* (ea >> 12) & 0xffff */ + xor r28,r5,r0 + + /* Convert linux PTE bits into HW equivalents */ + andi. r3,r30,0x1fe /* Get basic set of flags */ + xori r3,r3,HW_NO_EXEC /* _PAGE_EXEC -> NOEXEC */ + rlwinm r0,r30,32-9+1,30,30 /* _PAGE_RW -> _PAGE_USER (r0) */ + rlwinm r4,r30,32-7+1,30,30 /* _PAGE_DIRTY -> _PAGE_USER (r4) */ + and r0,r0,r4 /* _PAGE_RW & _PAGE_DIRTY -> r0 bit 30 */ + andc r0,r30,r0 /* r0 = pte & ~r0 */ + rlwimi r3,r0,32-1,31,31 /* Insert result into PP lsb */ + + /* We eventually do the icache sync here (maybe inline that + * code rather than call a C function...) + */ +BEGIN_FTR_SECTION +BEGIN_FTR_SECTION + mr r4,r30 + mr r5,r7 + bl .hash_page_do_lazy_icache +END_FTR_SECTION_IFSET(CPU_FTR_NOEXECUTE) +END_FTR_SECTION_IFCLR(CPU_FTR_COHERENT_ICACHE) + + /* At this point, r3 contains new PP bits, save them in + * place of "access" in the param area (sic) + */ + std r3,STK_PARM(r4)(r1) + + /* Get htab_hash_mask */ + ld r4,htab_hash_mask@got(2) + ld r27,0(r4) /* htab_hash_mask -> r27 */ + + /* Check if we may already be in the hashtable, in this case, we + * go to out-of-line code to try to modify the HPTE + */ + andi. r0,r31,_PAGE_HASHPTE + bne htab_modify_pte + +htab_insert_pte: + /* Clear hpte bits in new pte (we also clear BUSY btw) and + * add _PAGE_HASHPTE + */ + lis r0,_PAGE_HPTEFLAGS@h + ori r0,r0,_PAGE_HPTEFLAGS@l + andc r30,r30,r0 + ori r30,r30,_PAGE_HASHPTE + + /* page number in r5 */ + rldicl r5,r31,64-PTE_SHIFT,PTE_SHIFT + + /* Calculate primary group hash */ + and r0,r28,r27 + rldicr r3,r0,3,63-3 /* r0 = (hash & mask) << 3 */ + + /* Call ppc_md.hpte_insert */ + ld r7,STK_PARM(r4)(r1) /* Retreive new pp bits */ + mr r4,r29 /* Retreive va */ + li r6,0 /* primary slot */ + li r8,0 /* not bolted and not large */ + li r9,0 +_GLOBAL(htab_call_hpte_insert1) + bl . /* Will be patched by htab_finish_init() */ + cmpdi 0,r3,0 + bge htab_pte_insert_ok /* Insertion successful */ + cmpdi 0,r3,-2 /* Critical failure */ + beq- htab_pte_insert_failure + + /* Now try secondary slot */ + + /* page number in r5 */ + rldicl r5,r31,64-PTE_SHIFT,PTE_SHIFT + + /* Calculate secondary group hash */ + andc r0,r27,r28 + rldicr r3,r0,3,63-3 /* r0 = (~hash & mask) << 3 */ + + /* Call ppc_md.hpte_insert */ + ld r7,STK_PARM(r4)(r1) /* Retreive new pp bits */ + mr r4,r29 /* Retreive va */ + li r6,1 /* secondary slot */ + li r8,0 /* not bolted and not large */ + li r9,0 +_GLOBAL(htab_call_hpte_insert2) + bl . /* Will be patched by htab_finish_init() */ + cmpdi 0,r3,0 + bge+ htab_pte_insert_ok /* Insertion successful */ + cmpdi 0,r3,-2 /* Critical failure */ + beq- htab_pte_insert_failure + + /* Both are full, we need to evict something */ + mftb r0 + /* Pick a random group based on TB */ + andi. r0,r0,1 + mr r5,r28 + bne 2f + not r5,r5 +2: and r0,r5,r27 + rldicr r3,r0,3,63-3 /* r0 = (hash & mask) << 3 */ + /* Call ppc_md.hpte_remove */ +_GLOBAL(htab_call_hpte_remove) + bl . /* Will be patched by htab_finish_init() */ + + /* Try all again */ + b htab_insert_pte + +htab_pte_insert_ok: + /* Insert slot number & secondary bit in PTE */ + rldimi r30,r3,12,63-15 + + /* Write out the PTE with a normal write + * (maybe add eieio may be good still ?) + */ +htab_write_out_pte: + ld r6,STK_PARM(r6)(r1) + std r30,0(r6) + li r3, 0 +bail: + ld r27,STK_REG(r27)(r1) + ld r28,STK_REG(r28)(r1) + ld r29,STK_REG(r29)(r1) + ld r30,STK_REG(r30)(r1) + ld r31,STK_REG(r31)(r1) + addi r1,r1,STACKFRAMESIZE + ld r0,16(r1) + mtlr r0 + blr + +htab_modify_pte: + /* Keep PP bits in r4 and slot idx from the PTE around in r3 */ + mr r4,r3 + rlwinm r3,r31,32-12,29,31 + + /* Secondary group ? if yes, get a inverted hash value */ + mr r5,r28 + andi. r0,r31,_PAGE_SECONDARY + beq 1f + not r5,r5 +1: + /* Calculate proper slot value for ppc_md.hpte_updatepp */ + and r0,r5,r27 + rldicr r0,r0,3,63-3 /* r0 = (hash & mask) << 3 */ + add r3,r0,r3 /* add slot idx */ + + /* Call ppc_md.hpte_updatepp */ + mr r5,r29 /* va */ + li r6,0 /* large is 0 */ + ld r7,STK_PARM(r8)(r1) /* get "local" param */ +_GLOBAL(htab_call_hpte_updatepp) + bl . /* Will be patched by htab_finish_init() */ + + /* if we failed because typically the HPTE wasn't really here + * we try an insertion. + */ + cmpdi 0,r3,-1 + beq- htab_insert_pte + + /* Clear the BUSY bit and Write out the PTE */ + li r0,_PAGE_BUSY + andc r30,r30,r0 + b htab_write_out_pte + +htab_wrong_access: + /* Bail out clearing reservation */ + stdcx. r31,0,r6 + li r3,1 + b bail + +htab_pte_insert_failure: + /* Bail out restoring old PTE */ + ld r6,STK_PARM(r6)(r1) + std r31,0(r6) + li r3,-1 + b bail + + diff --git a/arch/ppc64/mm/hash_native.c b/arch/ppc64/mm/hash_native.c new file mode 100644 index 00000000000..144657e0c3d --- /dev/null +++ b/arch/ppc64/mm/hash_native.c @@ -0,0 +1,423 @@ +/* + * native hashtable management. + * + * SMP scalability work: + * Copyright (C) 2001 Anton Blanchard , IBM + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include + +#define HPTE_LOCK_BIT 3 + +static DEFINE_SPINLOCK(native_tlbie_lock); + +static inline void native_lock_hpte(HPTE *hptep) +{ + unsigned long *word = &hptep->dw0.dword0; + + while (1) { + if (!test_and_set_bit(HPTE_LOCK_BIT, word)) + break; + while(test_bit(HPTE_LOCK_BIT, word)) + cpu_relax(); + } +} + +static inline void native_unlock_hpte(HPTE *hptep) +{ + unsigned long *word = &hptep->dw0.dword0; + + asm volatile("lwsync":::"memory"); + clear_bit(HPTE_LOCK_BIT, word); +} + +long native_hpte_insert(unsigned long hpte_group, unsigned long va, + unsigned long prpn, int secondary, + unsigned long hpteflags, int bolted, int large) +{ + unsigned long arpn = physRpn_to_absRpn(prpn); + HPTE *hptep = htab_address + hpte_group; + Hpte_dword0 dw0; + HPTE lhpte; + int i; + + for (i = 0; i < HPTES_PER_GROUP; i++) { + dw0 = hptep->dw0.dw0; + + if (!dw0.v) { + /* retry with lock held */ + native_lock_hpte(hptep); + dw0 = hptep->dw0.dw0; + if (!dw0.v) + break; + native_unlock_hpte(hptep); + } + + hptep++; + } + + if (i == HPTES_PER_GROUP) + return -1; + + lhpte.dw1.dword1 = 0; + lhpte.dw1.dw1.rpn = arpn; + lhpte.dw1.flags.flags = hpteflags; + + lhpte.dw0.dword0 = 0; + lhpte.dw0.dw0.avpn = va >> 23; + lhpte.dw0.dw0.h = secondary; + lhpte.dw0.dw0.bolted = bolted; + lhpte.dw0.dw0.v = 1; + + if (large) { + lhpte.dw0.dw0.l = 1; + lhpte.dw0.dw0.avpn &= ~0x1UL; + } + + hptep->dw1.dword1 = lhpte.dw1.dword1; + + /* Guarantee the second dword is visible before the valid bit */ + __asm__ __volatile__ ("eieio" : : : "memory"); + + /* + * Now set the first dword including the valid bit + * NOTE: this also unlocks the hpte + */ + hptep->dw0.dword0 = lhpte.dw0.dword0; + + __asm__ __volatile__ ("ptesync" : : : "memory"); + + return i | (secondary << 3); +} + +static long native_hpte_remove(unsigned long hpte_group) +{ + HPTE *hptep; + Hpte_dword0 dw0; + int i; + int slot_offset; + + /* pick a random entry to start at */ + slot_offset = mftb() & 0x7; + + for (i = 0; i < HPTES_PER_GROUP; i++) { + hptep = htab_address + hpte_group + slot_offset; + dw0 = hptep->dw0.dw0; + + if (dw0.v && !dw0.bolted) { + /* retry with lock held */ + native_lock_hpte(hptep); + dw0 = hptep->dw0.dw0; + if (dw0.v && !dw0.bolted) + break; + native_unlock_hpte(hptep); + } + + slot_offset++; + slot_offset &= 0x7; + } + + if (i == HPTES_PER_GROUP) + return -1; + + /* Invalidate the hpte. NOTE: this also unlocks it */ + hptep->dw0.dword0 = 0; + + return i; +} + +static inline void set_pp_bit(unsigned long pp, HPTE *addr) +{ + unsigned long old; + unsigned long *p = &addr->dw1.dword1; + + __asm__ __volatile__( + "1: ldarx %0,0,%3\n\ + rldimi %0,%2,0,61\n\ + stdcx. %0,0,%3\n\ + bne 1b" + : "=&r" (old), "=m" (*p) + : "r" (pp), "r" (p), "m" (*p) + : "cc"); +} + +/* + * Only works on small pages. Yes its ugly to have to check each slot in + * the group but we only use this during bootup. + */ +static long native_hpte_find(unsigned long vpn) +{ + HPTE *hptep; + unsigned long hash; + unsigned long i, j; + long slot; + Hpte_dword0 dw0; + + hash = hpt_hash(vpn, 0); + + for (j = 0; j < 2; j++) { + slot = (hash & htab_hash_mask) * HPTES_PER_GROUP; + for (i = 0; i < HPTES_PER_GROUP; i++) { + hptep = htab_address + slot; + dw0 = hptep->dw0.dw0; + + if ((dw0.avpn == (vpn >> 11)) && dw0.v && + (dw0.h == j)) { + /* HPTE matches */ + if (j) + slot = -slot; + return slot; + } + ++slot; + } + hash = ~hash; + } + + return -1; +} + +static long native_hpte_updatepp(unsigned long slot, unsigned long newpp, + unsigned long va, int large, int local) +{ + HPTE *hptep = htab_address + slot; + Hpte_dword0 dw0; + unsigned long avpn = va >> 23; + int ret = 0; + + if (large) + avpn &= ~0x1UL; + + native_lock_hpte(hptep); + + dw0 = hptep->dw0.dw0; + + /* Even if we miss, we need to invalidate the TLB */ + if ((dw0.avpn != avpn) || !dw0.v) { + native_unlock_hpte(hptep); + ret = -1; + } else { + set_pp_bit(newpp, hptep); + native_unlock_hpte(hptep); + } + + /* Ensure it is out of the tlb too */ + if (cpu_has_feature(CPU_FTR_TLBIEL) && !large && local) { + tlbiel(va); + } else { + int lock_tlbie = !cpu_has_feature(CPU_FTR_LOCKLESS_TLBIE); + + if (lock_tlbie) + spin_lock(&native_tlbie_lock); + tlbie(va, large); + if (lock_tlbie) + spin_unlock(&native_tlbie_lock); + } + + return ret; +} + +/* + * Update the page protection bits. Intended to be used to create + * guard pages for kernel data structures on pages which are bolted + * in the HPT. Assumes pages being operated on will not be stolen. + * Does not work on large pages. + * + * No need to lock here because we should be the only user. + */ +static void native_hpte_updateboltedpp(unsigned long newpp, unsigned long ea) +{ + unsigned long vsid, va, vpn, flags = 0; + long slot; + HPTE *hptep; + int lock_tlbie = !cpu_has_feature(CPU_FTR_LOCKLESS_TLBIE); + + vsid = get_kernel_vsid(ea); + va = (vsid << 28) | (ea & 0x0fffffff); + vpn = va >> PAGE_SHIFT; + + slot = native_hpte_find(vpn); + if (slot == -1) + panic("could not find page to bolt\n"); + hptep = htab_address + slot; + + set_pp_bit(newpp, hptep); + + /* Ensure it is out of the tlb too */ + if (lock_tlbie) + spin_lock_irqsave(&native_tlbie_lock, flags); + tlbie(va, 0); + if (lock_tlbie) + spin_unlock_irqrestore(&native_tlbie_lock, flags); +} + +static void native_hpte_invalidate(unsigned long slot, unsigned long va, + int large, int local) +{ + HPTE *hptep = htab_address + slot; + Hpte_dword0 dw0; + unsigned long avpn = va >> 23; + unsigned long flags; + int lock_tlbie = !cpu_has_feature(CPU_FTR_LOCKLESS_TLBIE); + + if (large) + avpn &= ~0x1UL; + + local_irq_save(flags); + native_lock_hpte(hptep); + + dw0 = hptep->dw0.dw0; + + /* Even if we miss, we need to invalidate the TLB */ + if ((dw0.avpn != avpn) || !dw0.v) { + native_unlock_hpte(hptep); + } else { + /* Invalidate the hpte. NOTE: this also unlocks it */ + hptep->dw0.dword0 = 0; + } + + /* Invalidate the tlb */ + if (cpu_has_feature(CPU_FTR_TLBIEL) && !large && local) { + tlbiel(va); + } else { + if (lock_tlbie) + spin_lock(&native_tlbie_lock); + tlbie(va, large); + if (lock_tlbie) + spin_unlock(&native_tlbie_lock); + } + local_irq_restore(flags); +} + +static void native_flush_hash_range(unsigned long context, + unsigned long number, int local) +{ + unsigned long vsid, vpn, va, hash, secondary, slot, flags, avpn; + int i, j; + HPTE *hptep; + Hpte_dword0 dw0; + struct ppc64_tlb_batch *batch = &__get_cpu_var(ppc64_tlb_batch); + + /* XXX fix for large ptes */ + unsigned long large = 0; + + local_irq_save(flags); + + j = 0; + for (i = 0; i < number; i++) { + if ((batch->addr[i] >= USER_START) && + (batch->addr[i] <= USER_END)) + vsid = get_vsid(context, batch->addr[i]); + else + vsid = get_kernel_vsid(batch->addr[i]); + + va = (vsid << 28) | (batch->addr[i] & 0x0fffffff); + batch->vaddr[j] = va; + if (large) + vpn = va >> HPAGE_SHIFT; + else + vpn = va >> PAGE_SHIFT; + hash = hpt_hash(vpn, large); + secondary = (pte_val(batch->pte[i]) & _PAGE_SECONDARY) >> 15; + if (secondary) + hash = ~hash; + slot = (hash & htab_hash_mask) * HPTES_PER_GROUP; + slot += (pte_val(batch->pte[i]) & _PAGE_GROUP_IX) >> 12; + + hptep = htab_address + slot; + + avpn = va >> 23; + if (large) + avpn &= ~0x1UL; + + native_lock_hpte(hptep); + + dw0 = hptep->dw0.dw0; + + /* Even if we miss, we need to invalidate the TLB */ + if ((dw0.avpn != avpn) || !dw0.v) { + native_unlock_hpte(hptep); + } else { + /* Invalidate the hpte. NOTE: this also unlocks it */ + hptep->dw0.dword0 = 0; + } + + j++; + } + + if (cpu_has_feature(CPU_FTR_TLBIEL) && !large && local) { + asm volatile("ptesync":::"memory"); + + for (i = 0; i < j; i++) + __tlbiel(batch->vaddr[i]); + + asm volatile("ptesync":::"memory"); + } else { + int lock_tlbie = !cpu_has_feature(CPU_FTR_LOCKLESS_TLBIE); + + if (lock_tlbie) + spin_lock(&native_tlbie_lock); + + asm volatile("ptesync":::"memory"); + + for (i = 0; i < j; i++) + __tlbie(batch->vaddr[i], 0); + + asm volatile("eieio; tlbsync; ptesync":::"memory"); + + if (lock_tlbie) + spin_unlock(&native_tlbie_lock); + } + + local_irq_restore(flags); +} + +#ifdef CONFIG_PPC_PSERIES +/* Disable TLB batching on nighthawk */ +static inline int tlb_batching_enabled(void) +{ + struct device_node *root = of_find_node_by_path("/"); + int enabled = 1; + + if (root) { + const char *model = get_property(root, "model", NULL); + if (model && !strcmp(model, "IBM,9076-N81")) + enabled = 0; + of_node_put(root); + } + + return enabled; +} +#else +static inline int tlb_batching_enabled(void) +{ + return 1; +} +#endif + +void hpte_init_native(void) +{ + ppc_md.hpte_invalidate = native_hpte_invalidate; + ppc_md.hpte_updatepp = native_hpte_updatepp; + ppc_md.hpte_updateboltedpp = native_hpte_updateboltedpp; + ppc_md.hpte_insert = native_hpte_insert; + ppc_md.hpte_remove = native_hpte_remove; + if (tlb_batching_enabled()) + ppc_md.flush_hash_range = native_flush_hash_range; + htab_finish_init(); +} diff --git a/arch/ppc64/mm/hash_utils.c b/arch/ppc64/mm/hash_utils.c new file mode 100644 index 00000000000..e48be12f518 --- /dev/null +++ b/arch/ppc64/mm/hash_utils.c @@ -0,0 +1,439 @@ +/* + * PowerPC64 port by Mike Corrigan and Dave Engebretsen + * {mikejc|engebret}@us.ibm.com + * + * Copyright (c) 2000 Mike Corrigan + * + * SMP scalability work: + * Copyright (C) 2001 Anton Blanchard , IBM + * + * Module name: htab.c + * + * Description: + * PowerPC Hashed Page Table functions + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +#undef DEBUG + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#ifdef DEBUG +#define DBG(fmt...) udbg_printf(fmt) +#else +#define DBG(fmt...) +#endif + +/* + * Note: pte --> Linux PTE + * HPTE --> PowerPC Hashed Page Table Entry + * + * Execution context: + * htab_initialize is called with the MMU off (of course), but + * the kernel has been copied down to zero so it can directly + * reference global data. At this point it is very difficult + * to print debug info. + * + */ + +#ifdef CONFIG_U3_DART +extern unsigned long dart_tablebase; +#endif /* CONFIG_U3_DART */ + +HPTE *htab_address; +unsigned long htab_hash_mask; + +extern unsigned long _SDR1; + +#define KB (1024) +#define MB (1024*KB) + +static inline void loop_forever(void) +{ + volatile unsigned long x = 1; + for(;x;x|=1) + ; +} + +#ifdef CONFIG_PPC_MULTIPLATFORM +static inline void create_pte_mapping(unsigned long start, unsigned long end, + unsigned long mode, int large) +{ + unsigned long addr; + unsigned int step; + unsigned long tmp_mode; + + if (large) + step = 16*MB; + else + step = 4*KB; + + for (addr = start; addr < end; addr += step) { + unsigned long vpn, hash, hpteg; + unsigned long vsid = get_kernel_vsid(addr); + unsigned long va = (vsid << 28) | (addr & 0xfffffff); + int ret; + + if (large) + vpn = va >> HPAGE_SHIFT; + else + vpn = va >> PAGE_SHIFT; + + + tmp_mode = mode; + + /* Make non-kernel text non-executable */ + if (!in_kernel_text(addr)) + tmp_mode = mode | HW_NO_EXEC; + + hash = hpt_hash(vpn, large); + + hpteg = ((hash & htab_hash_mask) * HPTES_PER_GROUP); + +#ifdef CONFIG_PPC_PSERIES + if (systemcfg->platform & PLATFORM_LPAR) + ret = pSeries_lpar_hpte_insert(hpteg, va, + virt_to_abs(addr) >> PAGE_SHIFT, + 0, tmp_mode, 1, large); + else +#endif /* CONFIG_PPC_PSERIES */ + ret = native_hpte_insert(hpteg, va, + virt_to_abs(addr) >> PAGE_SHIFT, + 0, tmp_mode, 1, large); + + if (ret == -1) { + ppc64_terminate_msg(0x20, "create_pte_mapping"); + loop_forever(); + } + } +} + +void __init htab_initialize(void) +{ + unsigned long table, htab_size_bytes; + unsigned long pteg_count; + unsigned long mode_rw; + int i, use_largepages = 0; + unsigned long base = 0, size = 0; + extern unsigned long tce_alloc_start, tce_alloc_end; + + DBG(" -> htab_initialize()\n"); + + /* + * Calculate the required size of the htab. We want the number of + * PTEGs to equal one half the number of real pages. + */ + htab_size_bytes = 1UL << ppc64_pft_size; + pteg_count = htab_size_bytes >> 7; + + /* For debug, make the HTAB 1/8 as big as it normally would be. */ + ifppcdebug(PPCDBG_HTABSIZE) { + pteg_count >>= 3; + htab_size_bytes = pteg_count << 7; + } + + htab_hash_mask = pteg_count - 1; + + if (systemcfg->platform & PLATFORM_LPAR) { + /* Using a hypervisor which owns the htab */ + htab_address = NULL; + _SDR1 = 0; + } else { + /* Find storage for the HPT. Must be contiguous in + * the absolute address space. + */ + table = lmb_alloc(htab_size_bytes, htab_size_bytes); + + DBG("Hash table allocated at %lx, size: %lx\n", table, + htab_size_bytes); + + if ( !table ) { + ppc64_terminate_msg(0x20, "hpt space"); + loop_forever(); + } + htab_address = abs_to_virt(table); + + /* htab absolute addr + encoded htabsize */ + _SDR1 = table + __ilog2(pteg_count) - 11; + + /* Initialize the HPT with no entries */ + memset((void *)table, 0, htab_size_bytes); + } + + mode_rw = _PAGE_ACCESSED | _PAGE_COHERENT | PP_RWXX; + + /* On U3 based machines, we need to reserve the DART area and + * _NOT_ map it to avoid cache paradoxes as it's remapped non + * cacheable later on + */ + if (cpu_has_feature(CPU_FTR_16M_PAGE)) + use_largepages = 1; + + /* create bolted the linear mapping in the hash table */ + for (i=0; i < lmb.memory.cnt; i++) { + base = lmb.memory.region[i].physbase + KERNELBASE; + size = lmb.memory.region[i].size; + + DBG("creating mapping for region: %lx : %lx\n", base, size); + +#ifdef CONFIG_U3_DART + /* Do not map the DART space. Fortunately, it will be aligned + * in such a way that it will not cross two lmb regions and will + * fit within a single 16Mb page. + * The DART space is assumed to be a full 16Mb region even if we + * only use 2Mb of that space. We will use more of it later for + * AGP GART. We have to use a full 16Mb large page. + */ + DBG("DART base: %lx\n", dart_tablebase); + + if (dart_tablebase != 0 && dart_tablebase >= base + && dart_tablebase < (base + size)) { + if (base != dart_tablebase) + create_pte_mapping(base, dart_tablebase, mode_rw, + use_largepages); + if ((base + size) > (dart_tablebase + 16*MB)) + create_pte_mapping(dart_tablebase + 16*MB, base + size, + mode_rw, use_largepages); + continue; + } +#endif /* CONFIG_U3_DART */ + create_pte_mapping(base, base + size, mode_rw, use_largepages); + } + + /* + * If we have a memory_limit and we've allocated TCEs then we need to + * explicitly map the TCE area at the top of RAM. We also cope with the + * case that the TCEs start below memory_limit. + * tce_alloc_start/end are 16MB aligned so the mapping should work + * for either 4K or 16MB pages. + */ + if (tce_alloc_start) { + tce_alloc_start += KERNELBASE; + tce_alloc_end += KERNELBASE; + + if (base + size >= tce_alloc_start) + tce_alloc_start = base + size + 1; + + create_pte_mapping(tce_alloc_start, tce_alloc_end, + mode_rw, use_largepages); + } + + DBG(" <- htab_initialize()\n"); +} +#undef KB +#undef MB +#endif /* CONFIG_PPC_MULTIPLATFORM */ + +/* + * Called by asm hashtable.S for doing lazy icache flush + */ +unsigned int hash_page_do_lazy_icache(unsigned int pp, pte_t pte, int trap) +{ + struct page *page; + + if (!pfn_valid(pte_pfn(pte))) + return pp; + + page = pte_page(pte); + + /* page is dirty */ + if (!test_bit(PG_arch_1, &page->flags) && !PageReserved(page)) { + if (trap == 0x400) { + __flush_dcache_icache(page_address(page)); + set_bit(PG_arch_1, &page->flags); + } else + pp |= HW_NO_EXEC; + } + return pp; +} + +/* Result code is: + * 0 - handled + * 1 - normal page fault + * -1 - critical hash insertion error + */ +int hash_page(unsigned long ea, unsigned long access, unsigned long trap) +{ + void *pgdir; + unsigned long vsid; + struct mm_struct *mm; + pte_t *ptep; + int ret; + int user_region = 0; + int local = 0; + cpumask_t tmp; + + switch (REGION_ID(ea)) { + case USER_REGION_ID: + user_region = 1; + mm = current->mm; + if ((ea > USER_END) || (! mm)) + return 1; + + vsid = get_vsid(mm->context.id, ea); + break; + case IO_REGION_ID: + if (ea > IMALLOC_END) + return 1; + mm = &ioremap_mm; + vsid = get_kernel_vsid(ea); + break; + case VMALLOC_REGION_ID: + if (ea > VMALLOC_END) + return 1; + mm = &init_mm; + vsid = get_kernel_vsid(ea); + break; +#if 0 + case KERNEL_REGION_ID: + /* + * Should never get here - entire 0xC0... region is bolted. + * Send the problem up to do_page_fault + */ +#endif + default: + /* Not a valid range + * Send the problem up to do_page_fault + */ + return 1; + break; + } + + pgdir = mm->pgd; + + if (pgdir == NULL) + return 1; + + tmp = cpumask_of_cpu(smp_processor_id()); + if (user_region && cpus_equal(mm->cpu_vm_mask, tmp)) + local = 1; + + /* Is this a huge page ? */ + if (unlikely(in_hugepage_area(mm->context, ea))) + ret = hash_huge_page(mm, access, ea, vsid, local); + else { + ptep = find_linux_pte(pgdir, ea); + if (ptep == NULL) + return 1; + ret = __hash_page(ea, access, vsid, ptep, trap, local); + } + + return ret; +} + +void flush_hash_page(unsigned long context, unsigned long ea, pte_t pte, + int local) +{ + unsigned long vsid, vpn, va, hash, secondary, slot; + unsigned long huge = pte_huge(pte); + + if ((ea >= USER_START) && (ea <= USER_END)) + vsid = get_vsid(context, ea); + else + vsid = get_kernel_vsid(ea); + + va = (vsid << 28) | (ea & 0x0fffffff); + if (huge) + vpn = va >> HPAGE_SHIFT; + else + vpn = va >> PAGE_SHIFT; + hash = hpt_hash(vpn, huge); + secondary = (pte_val(pte) & _PAGE_SECONDARY) >> 15; + if (secondary) + hash = ~hash; + slot = (hash & htab_hash_mask) * HPTES_PER_GROUP; + slot += (pte_val(pte) & _PAGE_GROUP_IX) >> 12; + + ppc_md.hpte_invalidate(slot, va, huge, local); +} + +void flush_hash_range(unsigned long context, unsigned long number, int local) +{ + if (ppc_md.flush_hash_range) { + ppc_md.flush_hash_range(context, number, local); + } else { + int i; + struct ppc64_tlb_batch *batch = &__get_cpu_var(ppc64_tlb_batch); + + for (i = 0; i < number; i++) + flush_hash_page(context, batch->addr[i], batch->pte[i], + local); + } +} + +static inline void make_bl(unsigned int *insn_addr, void *func) +{ + unsigned long funcp = *((unsigned long *)func); + int offset = funcp - (unsigned long)insn_addr; + + *insn_addr = (unsigned int)(0x48000001 | (offset & 0x03fffffc)); + flush_icache_range((unsigned long)insn_addr, 4+ + (unsigned long)insn_addr); +} + +/* + * low_hash_fault is called when we the low level hash code failed + * to instert a PTE due to an hypervisor error + */ +void low_hash_fault(struct pt_regs *regs, unsigned long address) +{ + if (user_mode(regs)) { + siginfo_t info; + + info.si_signo = SIGBUS; + info.si_errno = 0; + info.si_code = BUS_ADRERR; + info.si_addr = (void __user *)address; + force_sig_info(SIGBUS, &info, current); + return; + } + bad_page_fault(regs, address, SIGBUS); +} + +void __init htab_finish_init(void) +{ + extern unsigned int *htab_call_hpte_insert1; + extern unsigned int *htab_call_hpte_insert2; + extern unsigned int *htab_call_hpte_remove; + extern unsigned int *htab_call_hpte_updatepp; + + make_bl(htab_call_hpte_insert1, ppc_md.hpte_insert); + make_bl(htab_call_hpte_insert2, ppc_md.hpte_insert); + make_bl(htab_call_hpte_remove, ppc_md.hpte_remove); + make_bl(htab_call_hpte_updatepp, ppc_md.hpte_updatepp); +} diff --git a/arch/ppc64/mm/hugetlbpage.c b/arch/ppc64/mm/hugetlbpage.c new file mode 100644 index 00000000000..c62ddaff072 --- /dev/null +++ b/arch/ppc64/mm/hugetlbpage.c @@ -0,0 +1,904 @@ +/* + * PPC64 (POWER4) Huge TLB Page Support for Kernel. + * + * Copyright (C) 2003 David Gibson, IBM Corporation. + * + * Based on the IA-32 version: + * Copyright (C) 2002, Rohit Seth + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +#define HUGEPGDIR_SHIFT (HPAGE_SHIFT + PAGE_SHIFT - 3) +#define HUGEPGDIR_SIZE (1UL << HUGEPGDIR_SHIFT) +#define HUGEPGDIR_MASK (~(HUGEPGDIR_SIZE-1)) + +#define HUGEPTE_INDEX_SIZE 9 +#define HUGEPGD_INDEX_SIZE 10 + +#define PTRS_PER_HUGEPTE (1 << HUGEPTE_INDEX_SIZE) +#define PTRS_PER_HUGEPGD (1 << HUGEPGD_INDEX_SIZE) + +static inline int hugepgd_index(unsigned long addr) +{ + return (addr & ~REGION_MASK) >> HUGEPGDIR_SHIFT; +} + +static pgd_t *hugepgd_offset(struct mm_struct *mm, unsigned long addr) +{ + int index; + + if (! mm->context.huge_pgdir) + return NULL; + + + index = hugepgd_index(addr); + BUG_ON(index >= PTRS_PER_HUGEPGD); + return mm->context.huge_pgdir + index; +} + +static inline pte_t *hugepte_offset(pgd_t *dir, unsigned long addr) +{ + int index; + + if (pgd_none(*dir)) + return NULL; + + index = (addr >> HPAGE_SHIFT) % PTRS_PER_HUGEPTE; + return (pte_t *)pgd_page(*dir) + index; +} + +static pgd_t *hugepgd_alloc(struct mm_struct *mm, unsigned long addr) +{ + BUG_ON(! in_hugepage_area(mm->context, addr)); + + if (! mm->context.huge_pgdir) { + pgd_t *new; + spin_unlock(&mm->page_table_lock); + /* Don't use pgd_alloc(), because we want __GFP_REPEAT */ + new = kmem_cache_alloc(zero_cache, GFP_KERNEL | __GFP_REPEAT); + BUG_ON(memcmp(new, empty_zero_page, PAGE_SIZE)); + spin_lock(&mm->page_table_lock); + + /* + * Because we dropped the lock, we should re-check the + * entry, as somebody else could have populated it.. + */ + if (mm->context.huge_pgdir) + pgd_free(new); + else + mm->context.huge_pgdir = new; + } + return hugepgd_offset(mm, addr); +} + +static pte_t *hugepte_alloc(struct mm_struct *mm, pgd_t *dir, + unsigned long addr) +{ + if (! pgd_present(*dir)) { + pte_t *new; + + spin_unlock(&mm->page_table_lock); + new = kmem_cache_alloc(zero_cache, GFP_KERNEL | __GFP_REPEAT); + BUG_ON(memcmp(new, empty_zero_page, PAGE_SIZE)); + spin_lock(&mm->page_table_lock); + /* + * Because we dropped the lock, we should re-check the + * entry, as somebody else could have populated it.. + */ + if (pgd_present(*dir)) { + if (new) + kmem_cache_free(zero_cache, new); + } else { + struct page *ptepage; + + if (! new) + return NULL; + ptepage = virt_to_page(new); + ptepage->mapping = (void *) mm; + ptepage->index = addr & HUGEPGDIR_MASK; + pgd_populate(mm, dir, new); + } + } + + return hugepte_offset(dir, addr); +} + +static pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr) +{ + pgd_t *pgd; + + BUG_ON(! in_hugepage_area(mm->context, addr)); + + pgd = hugepgd_offset(mm, addr); + if (! pgd) + return NULL; + + return hugepte_offset(pgd, addr); +} + +static pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr) +{ + pgd_t *pgd; + + BUG_ON(! in_hugepage_area(mm->context, addr)); + + pgd = hugepgd_alloc(mm, addr); + if (! pgd) + return NULL; + + return hugepte_alloc(mm, pgd, addr); +} + +static void set_huge_pte(struct mm_struct *mm, struct vm_area_struct *vma, + unsigned long addr, struct page *page, + pte_t *ptep, int write_access) +{ + pte_t entry; + + add_mm_counter(mm, rss, HPAGE_SIZE / PAGE_SIZE); + if (write_access) { + entry = + pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot))); + } else { + entry = pte_wrprotect(mk_pte(page, vma->vm_page_prot)); + } + entry = pte_mkyoung(entry); + entry = pte_mkhuge(entry); + + set_pte_at(mm, addr, ptep, entry); +} + +/* + * This function checks for proper alignment of input addr and len parameters. + */ +int is_aligned_hugepage_range(unsigned long addr, unsigned long len) +{ + if (len & ~HPAGE_MASK) + return -EINVAL; + if (addr & ~HPAGE_MASK) + return -EINVAL; + if (! (within_hugepage_low_range(addr, len) + || within_hugepage_high_range(addr, len)) ) + return -EINVAL; + return 0; +} + +static void flush_segments(void *parm) +{ + u16 segs = (unsigned long) parm; + unsigned long i; + + asm volatile("isync" : : : "memory"); + + for (i = 0; i < 16; i++) { + if (! (segs & (1U << i))) + continue; + asm volatile("slbie %0" : : "r" (i << SID_SHIFT)); + } + + asm volatile("isync" : : : "memory"); +} + +static int prepare_low_seg_for_htlb(struct mm_struct *mm, unsigned long seg) +{ + unsigned long start = seg << SID_SHIFT; + unsigned long end = (seg+1) << SID_SHIFT; + struct vm_area_struct *vma; + unsigned long addr; + struct mmu_gather *tlb; + + BUG_ON(seg >= 16); + + /* Check no VMAs are in the region */ + vma = find_vma(mm, start); + if (vma && (vma->vm_start < end)) + return -EBUSY; + + /* Clean up any leftover PTE pages in the region */ + spin_lock(&mm->page_table_lock); + tlb = tlb_gather_mmu(mm, 0); + for (addr = start; addr < end; addr += PMD_SIZE) { + pgd_t *pgd = pgd_offset(mm, addr); + pmd_t *pmd; + struct page *page; + pte_t *pte; + int i; + + if (pgd_none(*pgd)) + continue; + pmd = pmd_offset(pgd, addr); + if (!pmd || pmd_none(*pmd)) + continue; + if (pmd_bad(*pmd)) { + pmd_ERROR(*pmd); + pmd_clear(pmd); + continue; + } + pte = (pte_t *)pmd_page_kernel(*pmd); + /* No VMAs, so there should be no PTEs, check just in case. */ + for (i = 0; i < PTRS_PER_PTE; i++) { + BUG_ON(!pte_none(*pte)); + pte++; + } + page = pmd_page(*pmd); + pmd_clear(pmd); + mm->nr_ptes--; + dec_page_state(nr_page_table_pages); + pte_free_tlb(tlb, page); + } + tlb_finish_mmu(tlb, start, end); + spin_unlock(&mm->page_table_lock); + + return 0; +} + +static int open_low_hpage_segs(struct mm_struct *mm, u16 newsegs) +{ + unsigned long i; + + newsegs &= ~(mm->context.htlb_segs); + if (! newsegs) + return 0; /* The segments we want are already open */ + + for (i = 0; i < 16; i++) + if ((1 << i) & newsegs) + if (prepare_low_seg_for_htlb(mm, i) != 0) + return -EBUSY; + + mm->context.htlb_segs |= newsegs; + + /* update the paca copy of the context struct */ + get_paca()->context = mm->context; + + /* the context change must make it to memory before the flush, + * so that further SLB misses do the right thing. */ + mb(); + on_each_cpu(flush_segments, (void *)(unsigned long)newsegs, 0, 1); + + return 0; +} + +int prepare_hugepage_range(unsigned long addr, unsigned long len) +{ + if (within_hugepage_high_range(addr, len)) + return 0; + else if ((addr < 0x100000000UL) && ((addr+len) < 0x100000000UL)) { + int err; + /* Yes, we need both tests, in case addr+len overflows + * 64-bit arithmetic */ + err = open_low_hpage_segs(current->mm, + LOW_ESID_MASK(addr, len)); + if (err) + printk(KERN_DEBUG "prepare_hugepage_range(%lx, %lx)" + " failed (segs: 0x%04hx)\n", addr, len, + LOW_ESID_MASK(addr, len)); + return err; + } + + return -EINVAL; +} + +int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src, + struct vm_area_struct *vma) +{ + pte_t *src_pte, *dst_pte, entry; + struct page *ptepage; + unsigned long addr = vma->vm_start; + unsigned long end = vma->vm_end; + int err = -ENOMEM; + + while (addr < end) { + dst_pte = huge_pte_alloc(dst, addr); + if (!dst_pte) + goto out; + + src_pte = huge_pte_offset(src, addr); + entry = *src_pte; + + ptepage = pte_page(entry); + get_page(ptepage); + add_mm_counter(dst, rss, HPAGE_SIZE / PAGE_SIZE); + set_pte_at(dst, addr, dst_pte, entry); + + addr += HPAGE_SIZE; + } + + err = 0; + out: + return err; +} + +int +follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma, + struct page **pages, struct vm_area_struct **vmas, + unsigned long *position, int *length, int i) +{ + unsigned long vpfn, vaddr = *position; + int remainder = *length; + + WARN_ON(!is_vm_hugetlb_page(vma)); + + vpfn = vaddr/PAGE_SIZE; + while (vaddr < vma->vm_end && remainder) { + if (pages) { + pte_t *pte; + struct page *page; + + pte = huge_pte_offset(mm, vaddr); + + /* hugetlb should be locked, and hence, prefaulted */ + WARN_ON(!pte || pte_none(*pte)); + + page = &pte_page(*pte)[vpfn % (HPAGE_SIZE/PAGE_SIZE)]; + + WARN_ON(!PageCompound(page)); + + get_page(page); + pages[i] = page; + } + + if (vmas) + vmas[i] = vma; + + vaddr += PAGE_SIZE; + ++vpfn; + --remainder; + ++i; + } + + *length = remainder; + *position = vaddr; + + return i; +} + +struct page * +follow_huge_addr(struct mm_struct *mm, unsigned long address, int write) +{ + pte_t *ptep; + struct page *page; + + if (! in_hugepage_area(mm->context, address)) + return ERR_PTR(-EINVAL); + + ptep = huge_pte_offset(mm, address); + page = pte_page(*ptep); + if (page) + page += (address % HPAGE_SIZE) / PAGE_SIZE; + + return page; +} + +int pmd_huge(pmd_t pmd) +{ + return 0; +} + +struct page * +follow_huge_pmd(struct mm_struct *mm, unsigned long address, + pmd_t *pmd, int write) +{ + BUG(); + return NULL; +} + +void unmap_hugepage_range(struct vm_area_struct *vma, + unsigned long start, unsigned long end) +{ + struct mm_struct *mm = vma->vm_mm; + unsigned long addr; + pte_t *ptep; + struct page *page; + + WARN_ON(!is_vm_hugetlb_page(vma)); + BUG_ON((start % HPAGE_SIZE) != 0); + BUG_ON((end % HPAGE_SIZE) != 0); + + for (addr = start; addr < end; addr += HPAGE_SIZE) { + pte_t pte; + + ptep = huge_pte_offset(mm, addr); + if (!ptep || pte_none(*ptep)) + continue; + + pte = *ptep; + page = pte_page(pte); + pte_clear(mm, addr, ptep); + + put_page(page); + } + add_mm_counter(mm, rss, -((end - start) >> PAGE_SHIFT)); + flush_tlb_pending(); +} + +void hugetlb_free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *prev, + unsigned long start, unsigned long end) +{ + /* Because the huge pgtables are only 2 level, they can take + * at most around 4M, much less than one hugepage which the + * process is presumably entitled to use. So we don't bother + * freeing up the pagetables on unmap, and wait until + * destroy_context() to clean up the lot. */ +} + +int hugetlb_prefault(struct address_space *mapping, struct vm_area_struct *vma) +{ + struct mm_struct *mm = current->mm; + unsigned long addr; + int ret = 0; + + WARN_ON(!is_vm_hugetlb_page(vma)); + BUG_ON((vma->vm_start % HPAGE_SIZE) != 0); + BUG_ON((vma->vm_end % HPAGE_SIZE) != 0); + + spin_lock(&mm->page_table_lock); + for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) { + unsigned long idx; + pte_t *pte = huge_pte_alloc(mm, addr); + struct page *page; + + if (!pte) { + ret = -ENOMEM; + goto out; + } + if (! pte_none(*pte)) + continue; + + idx = ((addr - vma->vm_start) >> HPAGE_SHIFT) + + (vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT)); + page = find_get_page(mapping, idx); + if (!page) { + /* charge the fs quota first */ + if (hugetlb_get_quota(mapping)) { + ret = -ENOMEM; + goto out; + } + page = alloc_huge_page(); + if (!page) { + hugetlb_put_quota(mapping); + ret = -ENOMEM; + goto out; + } + ret = add_to_page_cache(page, mapping, idx, GFP_ATOMIC); + if (! ret) { + unlock_page(page); + } else { + hugetlb_put_quota(mapping); + free_huge_page(page); + goto out; + } + } + set_huge_pte(mm, vma, addr, page, pte, vma->vm_flags & VM_WRITE); + } +out: + spin_unlock(&mm->page_table_lock); + return ret; +} + +/* Because we have an exclusive hugepage region which lies within the + * normal user address space, we have to take special measures to make + * non-huge mmap()s evade the hugepage reserved regions. */ +unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr, + unsigned long len, unsigned long pgoff, + unsigned long flags) +{ + struct mm_struct *mm = current->mm; + struct vm_area_struct *vma; + unsigned long start_addr; + + if (len > TASK_SIZE) + return -ENOMEM; + + if (addr) { + addr = PAGE_ALIGN(addr); + vma = find_vma(mm, addr); + if (((TASK_SIZE - len) >= addr) + && (!vma || (addr+len) <= vma->vm_start) + && !is_hugepage_only_range(mm, addr,len)) + return addr; + } + start_addr = addr = mm->free_area_cache; + +full_search: + vma = find_vma(mm, addr); + while (TASK_SIZE - len >= addr) { + BUG_ON(vma && (addr >= vma->vm_end)); + + if (touches_hugepage_low_range(mm, addr, len)) { + addr = ALIGN(addr+1, 1<vm_start) { + /* + * Remember the place where we stopped the search: + */ + mm->free_area_cache = addr + len; + return addr; + } + addr = vma->vm_end; + vma = vma->vm_next; + } + + /* Make sure we didn't miss any holes */ + if (start_addr != TASK_UNMAPPED_BASE) { + start_addr = addr = TASK_UNMAPPED_BASE; + goto full_search; + } + return -ENOMEM; +} + +/* + * This mmap-allocator allocates new areas top-down from below the + * stack's low limit (the base): + * + * Because we have an exclusive hugepage region which lies within the + * normal user address space, we have to take special measures to make + * non-huge mmap()s evade the hugepage reserved regions. + */ +unsigned long +arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0, + const unsigned long len, const unsigned long pgoff, + const unsigned long flags) +{ + struct vm_area_struct *vma, *prev_vma; + struct mm_struct *mm = current->mm; + unsigned long base = mm->mmap_base, addr = addr0; + int first_time = 1; + + /* requested length too big for entire address space */ + if (len > TASK_SIZE) + return -ENOMEM; + + /* dont allow allocations above current base */ + if (mm->free_area_cache > base) + mm->free_area_cache = base; + + /* requesting a specific address */ + if (addr) { + addr = PAGE_ALIGN(addr); + vma = find_vma(mm, addr); + if (TASK_SIZE - len >= addr && + (!vma || addr + len <= vma->vm_start) + && !is_hugepage_only_range(mm, addr,len)) + return addr; + } + +try_again: + /* make sure it can fit in the remaining address space */ + if (mm->free_area_cache < len) + goto fail; + + /* either no address requested or cant fit in requested address hole */ + addr = (mm->free_area_cache - len) & PAGE_MASK; + do { +hugepage_recheck: + if (touches_hugepage_low_range(mm, addr, len)) { + addr = (addr & ((~0) << SID_SHIFT)) - len; + goto hugepage_recheck; + } else if (touches_hugepage_high_range(addr, len)) { + addr = TASK_HPAGE_BASE - len; + } + + /* + * Lookup failure means no vma is above this address, + * i.e. return with success: + */ + if (!(vma = find_vma_prev(mm, addr, &prev_vma))) + return addr; + + /* + * new region fits between prev_vma->vm_end and + * vma->vm_start, use it: + */ + if (addr+len <= vma->vm_start && + (!prev_vma || (addr >= prev_vma->vm_end))) + /* remember the address as a hint for next time */ + return (mm->free_area_cache = addr); + else + /* pull free_area_cache down to the first hole */ + if (mm->free_area_cache == vma->vm_end) + mm->free_area_cache = vma->vm_start; + + /* try just below the current vma->vm_start */ + addr = vma->vm_start-len; + } while (len <= vma->vm_start); + +fail: + /* + * if hint left us with no space for the requested + * mapping then try again: + */ + if (first_time) { + mm->free_area_cache = base; + first_time = 0; + goto try_again; + } + /* + * A failed mmap() very likely causes application failure, + * so fall back to the bottom-up function here. This scenario + * can happen with large stack limits and large mmap() + * allocations. + */ + mm->free_area_cache = TASK_UNMAPPED_BASE; + addr = arch_get_unmapped_area(filp, addr0, len, pgoff, flags); + /* + * Restore the topdown base: + */ + mm->free_area_cache = base; + + return addr; +} + +static unsigned long htlb_get_low_area(unsigned long len, u16 segmask) +{ + unsigned long addr = 0; + struct vm_area_struct *vma; + + vma = find_vma(current->mm, addr); + while (addr + len <= 0x100000000UL) { + BUG_ON(vma && (addr >= vma->vm_end)); /* invariant */ + + if (! __within_hugepage_low_range(addr, len, segmask)) { + addr = ALIGN(addr+1, 1<mm, addr); + continue; + } + + if (!vma || (addr + len) <= vma->vm_start) + return addr; + addr = ALIGN(vma->vm_end, HPAGE_SIZE); + /* Depending on segmask this might not be a confirmed + * hugepage region, so the ALIGN could have skipped + * some VMAs */ + vma = find_vma(current->mm, addr); + } + + return -ENOMEM; +} + +static unsigned long htlb_get_high_area(unsigned long len) +{ + unsigned long addr = TASK_HPAGE_BASE; + struct vm_area_struct *vma; + + vma = find_vma(current->mm, addr); + for (vma = find_vma(current->mm, addr); + addr + len <= TASK_HPAGE_END; + vma = vma->vm_next) { + BUG_ON(vma && (addr >= vma->vm_end)); /* invariant */ + BUG_ON(! within_hugepage_high_range(addr, len)); + + if (!vma || (addr + len) <= vma->vm_start) + return addr; + addr = ALIGN(vma->vm_end, HPAGE_SIZE); + /* Because we're in a hugepage region, this alignment + * should not skip us over any VMAs */ + } + + return -ENOMEM; +} + +unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr, + unsigned long len, unsigned long pgoff, + unsigned long flags) +{ + if (len & ~HPAGE_MASK) + return -EINVAL; + + if (!cpu_has_feature(CPU_FTR_16M_PAGE)) + return -EINVAL; + + if (test_thread_flag(TIF_32BIT)) { + int lastshift = 0; + u16 segmask, cursegs = current->mm->context.htlb_segs; + + /* First see if we can do the mapping in the existing + * low hpage segments */ + addr = htlb_get_low_area(len, cursegs); + if (addr != -ENOMEM) + return addr; + + for (segmask = LOW_ESID_MASK(0x100000000UL-len, len); + ! lastshift; segmask >>=1) { + if (segmask & 1) + lastshift = 1; + + addr = htlb_get_low_area(len, cursegs | segmask); + if ((addr != -ENOMEM) + && open_low_hpage_segs(current->mm, segmask) == 0) + return addr; + } + printk(KERN_DEBUG "hugetlb_get_unmapped_area() unable to open" + " enough segments\n"); + return -ENOMEM; + } else { + return htlb_get_high_area(len); + } +} + +void hugetlb_mm_free_pgd(struct mm_struct *mm) +{ + int i; + pgd_t *pgdir; + + spin_lock(&mm->page_table_lock); + + pgdir = mm->context.huge_pgdir; + if (! pgdir) + goto out; + + mm->context.huge_pgdir = NULL; + + /* cleanup any hugepte pages leftover */ + for (i = 0; i < PTRS_PER_HUGEPGD; i++) { + pgd_t *pgd = pgdir + i; + + if (! pgd_none(*pgd)) { + pte_t *pte = (pte_t *)pgd_page(*pgd); + struct page *ptepage = virt_to_page(pte); + + ptepage->mapping = NULL; + + BUG_ON(memcmp(pte, empty_zero_page, PAGE_SIZE)); + kmem_cache_free(zero_cache, pte); + } + pgd_clear(pgd); + } + + BUG_ON(memcmp(pgdir, empty_zero_page, PAGE_SIZE)); + kmem_cache_free(zero_cache, pgdir); + + out: + spin_unlock(&mm->page_table_lock); +} + +int hash_huge_page(struct mm_struct *mm, unsigned long access, + unsigned long ea, unsigned long vsid, int local) +{ + pte_t *ptep; + unsigned long va, vpn; + pte_t old_pte, new_pte; + unsigned long hpteflags, prpn; + long slot; + int err = 1; + + spin_lock(&mm->page_table_lock); + + ptep = huge_pte_offset(mm, ea); + + /* Search the Linux page table for a match with va */ + va = (vsid << 28) | (ea & 0x0fffffff); + vpn = va >> HPAGE_SHIFT; + + /* + * If no pte found or not present, send the problem up to + * do_page_fault + */ + if (unlikely(!ptep || pte_none(*ptep))) + goto out; + +/* BUG_ON(pte_bad(*ptep)); */ + + /* + * Check the user's access rights to the page. If access should be + * prevented then send the problem up to do_page_fault. + */ + if (unlikely(access & ~pte_val(*ptep))) + goto out; + /* + * At this point, we have a pte (old_pte) which can be used to build + * or update an HPTE. There are 2 cases: + * + * 1. There is a valid (present) pte with no associated HPTE (this is + * the most common case) + * 2. There is a valid (present) pte with an associated HPTE. The + * current values of the pp bits in the HPTE prevent access + * because we are doing software DIRTY bit management and the + * page is currently not DIRTY. + */ + + + old_pte = *ptep; + new_pte = old_pte; + + hpteflags = 0x2 | (! (pte_val(new_pte) & _PAGE_RW)); + /* _PAGE_EXEC -> HW_NO_EXEC since it's inverted */ + hpteflags |= ((pte_val(new_pte) & _PAGE_EXEC) ? 0 : HW_NO_EXEC); + + /* Check if pte already has an hpte (case 2) */ + if (unlikely(pte_val(old_pte) & _PAGE_HASHPTE)) { + /* There MIGHT be an HPTE for this pte */ + unsigned long hash, slot; + + hash = hpt_hash(vpn, 1); + if (pte_val(old_pte) & _PAGE_SECONDARY) + hash = ~hash; + slot = (hash & htab_hash_mask) * HPTES_PER_GROUP; + slot += (pte_val(old_pte) & _PAGE_GROUP_IX) >> 12; + + if (ppc_md.hpte_updatepp(slot, hpteflags, va, 1, local) == -1) + pte_val(old_pte) &= ~_PAGE_HPTEFLAGS; + } + + if (likely(!(pte_val(old_pte) & _PAGE_HASHPTE))) { + unsigned long hash = hpt_hash(vpn, 1); + unsigned long hpte_group; + + prpn = pte_pfn(old_pte); + +repeat: + hpte_group = ((hash & htab_hash_mask) * + HPTES_PER_GROUP) & ~0x7UL; + + /* Update the linux pte with the HPTE slot */ + pte_val(new_pte) &= ~_PAGE_HPTEFLAGS; + pte_val(new_pte) |= _PAGE_HASHPTE; + + /* Add in WIMG bits */ + /* XXX We should store these in the pte */ + hpteflags |= _PAGE_COHERENT; + + slot = ppc_md.hpte_insert(hpte_group, va, prpn, 0, + hpteflags, 0, 1); + + /* Primary is full, try the secondary */ + if (unlikely(slot == -1)) { + pte_val(new_pte) |= _PAGE_SECONDARY; + hpte_group = ((~hash & htab_hash_mask) * + HPTES_PER_GROUP) & ~0x7UL; + slot = ppc_md.hpte_insert(hpte_group, va, prpn, + 1, hpteflags, 0, 1); + if (slot == -1) { + if (mftb() & 0x1) + hpte_group = ((hash & htab_hash_mask) * HPTES_PER_GROUP) & ~0x7UL; + + ppc_md.hpte_remove(hpte_group); + goto repeat; + } + } + + if (unlikely(slot == -2)) + panic("hash_huge_page: pte_insert failed\n"); + + pte_val(new_pte) |= (slot<<12) & _PAGE_GROUP_IX; + + /* + * No need to use ldarx/stdcx here because all who + * might be updating the pte will hold the + * page_table_lock + */ + *ptep = new_pte; + } + + err = 0; + + out: + spin_unlock(&mm->page_table_lock); + + return err; +} diff --git a/arch/ppc64/mm/imalloc.c b/arch/ppc64/mm/imalloc.c new file mode 100644 index 00000000000..9d92b0d9cde --- /dev/null +++ b/arch/ppc64/mm/imalloc.c @@ -0,0 +1,312 @@ +/* + * c 2001 PPC 64 Team, IBM Corp + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +#include +#include + +#include +#include +#include +#include + +static DECLARE_MUTEX(imlist_sem); +struct vm_struct * imlist = NULL; + +static int get_free_im_addr(unsigned long size, unsigned long *im_addr) +{ + unsigned long addr; + struct vm_struct **p, *tmp; + + addr = IMALLOC_START; + for (p = &imlist; (tmp = *p) ; p = &tmp->next) { + if (size + addr < (unsigned long) tmp->addr) + break; + if ((unsigned long)tmp->addr >= IMALLOC_START) + addr = tmp->size + (unsigned long) tmp->addr; + if (addr > IMALLOC_END-size) + return 1; + } + *im_addr = addr; + + return 0; +} + +/* Return whether the region described by v_addr and size is a subset + * of the region described by parent + */ +static inline int im_region_is_subset(unsigned long v_addr, unsigned long size, + struct vm_struct *parent) +{ + return (int) (v_addr >= (unsigned long) parent->addr && + v_addr < (unsigned long) parent->addr + parent->size && + size < parent->size); +} + +/* Return whether the region described by v_addr and size is a superset + * of the region described by child + */ +static int im_region_is_superset(unsigned long v_addr, unsigned long size, + struct vm_struct *child) +{ + struct vm_struct parent; + + parent.addr = (void *) v_addr; + parent.size = size; + + return im_region_is_subset((unsigned long) child->addr, child->size, + &parent); +} + +/* Return whether the region described by v_addr and size overlaps + * the region described by vm. Overlapping regions meet the + * following conditions: + * 1) The regions share some part of the address space + * 2) The regions aren't identical + * 3) Neither region is a subset of the other + */ +static int im_region_overlaps(unsigned long v_addr, unsigned long size, + struct vm_struct *vm) +{ + if (im_region_is_superset(v_addr, size, vm)) + return 0; + + return (v_addr + size > (unsigned long) vm->addr + vm->size && + v_addr < (unsigned long) vm->addr + vm->size) || + (v_addr < (unsigned long) vm->addr && + v_addr + size > (unsigned long) vm->addr); +} + +/* Determine imalloc status of region described by v_addr and size. + * Can return one of the following: + * IM_REGION_UNUSED - Entire region is unallocated in imalloc space. + * IM_REGION_SUBSET - Region is a subset of a region that is already + * allocated in imalloc space. + * vm will be assigned to a ptr to the parent region. + * IM_REGION_EXISTS - Exact region already allocated in imalloc space. + * vm will be assigned to a ptr to the existing imlist + * member. + * IM_REGION_OVERLAPS - Region overlaps an allocated region in imalloc space. + * IM_REGION_SUPERSET - Region is a superset of a region that is already + * allocated in imalloc space. + */ +static int im_region_status(unsigned long v_addr, unsigned long size, + struct vm_struct **vm) +{ + struct vm_struct *tmp; + + for (tmp = imlist; tmp; tmp = tmp->next) + if (v_addr < (unsigned long) tmp->addr + tmp->size) + break; + + if (tmp) { + if (im_region_overlaps(v_addr, size, tmp)) + return IM_REGION_OVERLAP; + + *vm = tmp; + if (im_region_is_subset(v_addr, size, tmp)) { + /* Return with tmp pointing to superset */ + return IM_REGION_SUBSET; + } + if (im_region_is_superset(v_addr, size, tmp)) { + /* Return with tmp pointing to first subset */ + return IM_REGION_SUPERSET; + } + else if (v_addr == (unsigned long) tmp->addr && + size == tmp->size) { + /* Return with tmp pointing to exact region */ + return IM_REGION_EXISTS; + } + } + + *vm = NULL; + return IM_REGION_UNUSED; +} + +static struct vm_struct * split_im_region(unsigned long v_addr, + unsigned long size, struct vm_struct *parent) +{ + struct vm_struct *vm1 = NULL; + struct vm_struct *vm2 = NULL; + struct vm_struct *new_vm = NULL; + + vm1 = (struct vm_struct *) kmalloc(sizeof(*vm1), GFP_KERNEL); + if (vm1 == NULL) { + printk(KERN_ERR "%s() out of memory\n", __FUNCTION__); + return NULL; + } + + if (v_addr == (unsigned long) parent->addr) { + /* Use existing parent vm_struct to represent child, allocate + * new one for the remainder of parent range + */ + vm1->size = parent->size - size; + vm1->addr = (void *) (v_addr + size); + vm1->next = parent->next; + + parent->size = size; + parent->next = vm1; + new_vm = parent; + } else if (v_addr + size == (unsigned long) parent->addr + + parent->size) { + /* Allocate new vm_struct to represent child, use existing + * parent one for remainder of parent range + */ + vm1->size = size; + vm1->addr = (void *) v_addr; + vm1->next = parent->next; + new_vm = vm1; + + parent->size -= size; + parent->next = vm1; + } else { + /* Allocate two new vm_structs for the new child and + * uppermost remainder, and use existing parent one for the + * lower remainder of parent range + */ + vm2 = (struct vm_struct *) kmalloc(sizeof(*vm2), GFP_KERNEL); + if (vm2 == NULL) { + printk(KERN_ERR "%s() out of memory\n", __FUNCTION__); + kfree(vm1); + return NULL; + } + + vm1->size = size; + vm1->addr = (void *) v_addr; + vm1->next = vm2; + new_vm = vm1; + + vm2->size = ((unsigned long) parent->addr + parent->size) - + (v_addr + size); + vm2->addr = (void *) v_addr + size; + vm2->next = parent->next; + + parent->size = v_addr - (unsigned long) parent->addr; + parent->next = vm1; + } + + return new_vm; +} + +static struct vm_struct * __add_new_im_area(unsigned long req_addr, + unsigned long size) +{ + struct vm_struct **p, *tmp, *area; + + for (p = &imlist; (tmp = *p) ; p = &tmp->next) { + if (req_addr + size <= (unsigned long)tmp->addr) + break; + } + + area = (struct vm_struct *) kmalloc(sizeof(*area), GFP_KERNEL); + if (!area) + return NULL; + area->flags = 0; + area->addr = (void *)req_addr; + area->size = size; + area->next = *p; + *p = area; + + return area; +} + +static struct vm_struct * __im_get_area(unsigned long req_addr, + unsigned long size, + int criteria) +{ + struct vm_struct *tmp; + int status; + + status = im_region_status(req_addr, size, &tmp); + if ((criteria & status) == 0) { + return NULL; + } + + switch (status) { + case IM_REGION_UNUSED: + tmp = __add_new_im_area(req_addr, size); + break; + case IM_REGION_SUBSET: + tmp = split_im_region(req_addr, size, tmp); + break; + case IM_REGION_EXISTS: + /* Return requested region */ + break; + case IM_REGION_SUPERSET: + /* Return first existing subset of requested region */ + break; + default: + printk(KERN_ERR "%s() unexpected imalloc region status\n", + __FUNCTION__); + tmp = NULL; + } + + return tmp; +} + +struct vm_struct * im_get_free_area(unsigned long size) +{ + struct vm_struct *area; + unsigned long addr; + + down(&imlist_sem); + if (get_free_im_addr(size, &addr)) { + printk(KERN_ERR "%s() cannot obtain addr for size 0x%lx\n", + __FUNCTION__, size); + area = NULL; + goto next_im_done; + } + + area = __im_get_area(addr, size, IM_REGION_UNUSED); + if (area == NULL) { + printk(KERN_ERR + "%s() cannot obtain area for addr 0x%lx size 0x%lx\n", + __FUNCTION__, addr, size); + } +next_im_done: + up(&imlist_sem); + return area; +} + +struct vm_struct * im_get_area(unsigned long v_addr, unsigned long size, + int criteria) +{ + struct vm_struct *area; + + down(&imlist_sem); + area = __im_get_area(v_addr, size, criteria); + up(&imlist_sem); + return area; +} + +unsigned long im_free(void * addr) +{ + struct vm_struct **p, *tmp; + unsigned long ret_size = 0; + + if (!addr) + return ret_size; + if ((PAGE_SIZE-1) & (unsigned long) addr) { + printk(KERN_ERR "Trying to %s bad address (%p)\n", __FUNCTION__, addr); + return ret_size; + } + down(&imlist_sem); + for (p = &imlist ; (tmp = *p) ; p = &tmp->next) { + if (tmp->addr == addr) { + ret_size = tmp->size; + *p = tmp->next; + kfree(tmp); + up(&imlist_sem); + return ret_size; + } + } + up(&imlist_sem); + printk(KERN_ERR "Trying to %s nonexistent area (%p)\n", __FUNCTION__, + addr); + return ret_size; +} diff --git a/arch/ppc64/mm/init.c b/arch/ppc64/mm/init.c new file mode 100644 index 00000000000..23813d03e1c --- /dev/null +++ b/arch/ppc64/mm/init.c @@ -0,0 +1,927 @@ +/* + * PowerPC version + * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org) + * + * Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au) + * and Cort Dougan (PReP) (cort@cs.nmt.edu) + * Copyright (C) 1996 Paul Mackerras + * Amiga/APUS changes by Jesper Skov (jskov@cygnus.co.uk). + * + * Derived from "arch/i386/mm/init.c" + * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds + * + * Dave Engebretsen + * Rework for PPC64 port. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + * + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +int mem_init_done; +unsigned long ioremap_bot = IMALLOC_BASE; +static unsigned long phbs_io_bot = PHBS_IO_BASE; + +extern pgd_t swapper_pg_dir[]; +extern struct task_struct *current_set[NR_CPUS]; + +extern pgd_t ioremap_dir[]; +pgd_t * ioremap_pgd = (pgd_t *)&ioremap_dir; + +unsigned long klimit = (unsigned long)_end; + +unsigned long _SDR1=0; +unsigned long _ASR=0; + +/* max amount of RAM to use */ +unsigned long __max_memory; + +/* info on what we think the IO hole is */ +unsigned long io_hole_start; +unsigned long io_hole_size; + +void show_mem(void) +{ + unsigned long total = 0, reserved = 0; + unsigned long shared = 0, cached = 0; + struct page *page; + pg_data_t *pgdat; + unsigned long i; + + printk("Mem-info:\n"); + show_free_areas(); + printk("Free swap: %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10)); + for_each_pgdat(pgdat) { + for (i = 0; i < pgdat->node_spanned_pages; i++) { + page = pgdat->node_mem_map + i; + total++; + if (PageReserved(page)) + reserved++; + else if (PageSwapCache(page)) + cached++; + else if (page_count(page)) + shared += page_count(page) - 1; + } + } + printk("%ld pages of RAM\n", total); + printk("%ld reserved pages\n", reserved); + printk("%ld pages shared\n", shared); + printk("%ld pages swap cached\n", cached); +} + +#ifdef CONFIG_PPC_ISERIES + +void __iomem *ioremap(unsigned long addr, unsigned long size) +{ + return (void __iomem *)addr; +} + +extern void __iomem *__ioremap(unsigned long addr, unsigned long size, + unsigned long flags) +{ + return (void __iomem *)addr; +} + +void iounmap(volatile void __iomem *addr) +{ + return; +} + +#else + +/* + * map_io_page currently only called by __ioremap + * map_io_page adds an entry to the ioremap page table + * and adds an entry to the HPT, possibly bolting it + */ +static void map_io_page(unsigned long ea, unsigned long pa, int flags) +{ + pgd_t *pgdp; + pmd_t *pmdp; + pte_t *ptep; + unsigned long vsid; + + if (mem_init_done) { + spin_lock(&ioremap_mm.page_table_lock); + pgdp = pgd_offset_i(ea); + pmdp = pmd_alloc(&ioremap_mm, pgdp, ea); + ptep = pte_alloc_kernel(&ioremap_mm, pmdp, ea); + + pa = abs_to_phys(pa); + set_pte_at(&ioremap_mm, ea, ptep, pfn_pte(pa >> PAGE_SHIFT, __pgprot(flags))); + spin_unlock(&ioremap_mm.page_table_lock); + } else { + unsigned long va, vpn, hash, hpteg; + + /* + * If the mm subsystem is not fully up, we cannot create a + * linux page table entry for this mapping. Simply bolt an + * entry in the hardware page table. + */ + vsid = get_kernel_vsid(ea); + va = (vsid << 28) | (ea & 0xFFFFFFF); + vpn = va >> PAGE_SHIFT; + + hash = hpt_hash(vpn, 0); + + hpteg = ((hash & htab_hash_mask) * HPTES_PER_GROUP); + + /* Panic if a pte grpup is full */ + if (ppc_md.hpte_insert(hpteg, va, pa >> PAGE_SHIFT, 0, + _PAGE_NO_CACHE|_PAGE_GUARDED|PP_RWXX, + 1, 0) == -1) { + panic("map_io_page: could not insert mapping"); + } + } +} + + +static void __iomem * __ioremap_com(unsigned long addr, unsigned long pa, + unsigned long ea, unsigned long size, + unsigned long flags) +{ + unsigned long i; + + if ((flags & _PAGE_PRESENT) == 0) + flags |= pgprot_val(PAGE_KERNEL); + if (flags & (_PAGE_NO_CACHE | _PAGE_WRITETHRU)) + flags |= _PAGE_GUARDED; + + for (i = 0; i < size; i += PAGE_SIZE) { + map_io_page(ea+i, pa+i, flags); + } + + return (void __iomem *) (ea + (addr & ~PAGE_MASK)); +} + + +void __iomem * +ioremap(unsigned long addr, unsigned long size) +{ + return __ioremap(addr, size, _PAGE_NO_CACHE); +} + +void __iomem * +__ioremap(unsigned long addr, unsigned long size, unsigned long flags) +{ + unsigned long pa, ea; + + /* + * Choose an address to map it to. + * Once the imalloc system is running, we use it. + * Before that, we map using addresses going + * up from ioremap_bot. imalloc will use + * the addresses from ioremap_bot through + * IMALLOC_END (0xE000001fffffffff) + * + */ + pa = addr & PAGE_MASK; + size = PAGE_ALIGN(addr + size) - pa; + + if (size == 0) + return NULL; + + if (mem_init_done) { + struct vm_struct *area; + area = im_get_free_area(size); + if (area == NULL) + return NULL; + ea = (unsigned long)(area->addr); + } else { + ea = ioremap_bot; + ioremap_bot += size; + } + + return __ioremap_com(addr, pa, ea, size, flags); +} + +#define IS_PAGE_ALIGNED(_val) ((_val) == ((_val) & PAGE_MASK)) + +int __ioremap_explicit(unsigned long pa, unsigned long ea, + unsigned long size, unsigned long flags) +{ + struct vm_struct *area; + + /* For now, require page-aligned values for pa, ea, and size */ + if (!IS_PAGE_ALIGNED(pa) || !IS_PAGE_ALIGNED(ea) || + !IS_PAGE_ALIGNED(size)) { + printk(KERN_ERR "unaligned value in %s\n", __FUNCTION__); + return 1; + } + + if (!mem_init_done) { + /* Two things to consider in this case: + * 1) No records will be kept (imalloc, etc) that the region + * has been remapped + * 2) It won't be easy to iounmap() the region later (because + * of 1) + */ + ; + } else { + area = im_get_area(ea, size, + IM_REGION_UNUSED|IM_REGION_SUBSET|IM_REGION_EXISTS); + if (area == NULL) { + /* Expected when PHB-dlpar is in play */ + return 1; + } + if (ea != (unsigned long) area->addr) { + printk(KERN_ERR "unexpected addr return from im_get_area\n"); + return 1; + } + } + + if (__ioremap_com(pa, pa, ea, size, flags) != (void *) ea) { + printk(KERN_ERR "__ioremap_com() returned unexpected addr\n"); + return 1; + } + + return 0; +} + +static void unmap_im_area_pte(pmd_t *pmd, unsigned long address, + unsigned long size) +{ + unsigned long base, end; + pte_t *pte; + + if (pmd_none(*pmd)) + return; + if (pmd_bad(*pmd)) { + pmd_ERROR(*pmd); + pmd_clear(pmd); + return; + } + + pte = pte_offset_kernel(pmd, address); + base = address & PMD_MASK; + address &= ~PMD_MASK; + end = address + size; + if (end > PMD_SIZE) + end = PMD_SIZE; + + do { + pte_t page; + page = ptep_get_and_clear(&ioremap_mm, base + address, pte); + address += PAGE_SIZE; + pte++; + if (pte_none(page)) + continue; + if (pte_present(page)) + continue; + printk(KERN_CRIT "Whee.. Swapped out page in kernel page table\n"); + } while (address < end); +} + +static void unmap_im_area_pmd(pgd_t *dir, unsigned long address, + unsigned long size) +{ + unsigned long base, end; + pmd_t *pmd; + + if (pgd_none(*dir)) + return; + if (pgd_bad(*dir)) { + pgd_ERROR(*dir); + pgd_clear(dir); + return; + } + + pmd = pmd_offset(dir, address); + base = address & PGDIR_MASK; + address &= ~PGDIR_MASK; + end = address + size; + if (end > PGDIR_SIZE) + end = PGDIR_SIZE; + + do { + unmap_im_area_pte(pmd, base + address, end - address); + address = (address + PMD_SIZE) & PMD_MASK; + pmd++; + } while (address < end); +} + +/* + * Unmap an IO region and remove it from imalloc'd list. + * Access to IO memory should be serialized by driver. + * This code is modeled after vmalloc code - unmap_vm_area() + * + * XXX what about calls before mem_init_done (ie python_countermeasures()) + */ +void iounmap(volatile void __iomem *token) +{ + unsigned long address, start, end, size; + struct mm_struct *mm; + pgd_t *dir; + void *addr; + + if (!mem_init_done) { + return; + } + + addr = (void *) ((unsigned long __force) token & PAGE_MASK); + + if ((size = im_free(addr)) == 0) { + return; + } + + address = (unsigned long)addr; + start = address; + end = address + size; + + mm = &ioremap_mm; + spin_lock(&mm->page_table_lock); + + dir = pgd_offset_i(address); + flush_cache_vunmap(address, end); + do { + unmap_im_area_pmd(dir, address, end - address); + address = (address + PGDIR_SIZE) & PGDIR_MASK; + dir++; + } while (address && (address < end)); + flush_tlb_kernel_range(start, end); + + spin_unlock(&mm->page_table_lock); + return; +} + +static int iounmap_subset_regions(unsigned long addr, unsigned long size) +{ + struct vm_struct *area; + + /* Check whether subsets of this region exist */ + area = im_get_area(addr, size, IM_REGION_SUPERSET); + if (area == NULL) + return 1; + + while (area) { + iounmap((void __iomem *) area->addr); + area = im_get_area(addr, size, + IM_REGION_SUPERSET); + } + + return 0; +} + +int iounmap_explicit(volatile void __iomem *start, unsigned long size) +{ + struct vm_struct *area; + unsigned long addr; + int rc; + + addr = (unsigned long __force) start & PAGE_MASK; + + /* Verify that the region either exists or is a subset of an existing + * region. In the latter case, split the parent region to create + * the exact region + */ + area = im_get_area(addr, size, + IM_REGION_EXISTS | IM_REGION_SUBSET); + if (area == NULL) { + /* Determine whether subset regions exist. If so, unmap */ + rc = iounmap_subset_regions(addr, size); + if (rc) { + printk(KERN_ERR + "%s() cannot unmap nonexistent range 0x%lx\n", + __FUNCTION__, addr); + return 1; + } + } else { + iounmap((void __iomem *) area->addr); + } + /* + * FIXME! This can't be right: + iounmap(area->addr); + * Maybe it should be "iounmap(area);" + */ + return 0; +} + +#endif + +EXPORT_SYMBOL(ioremap); +EXPORT_SYMBOL(__ioremap); +EXPORT_SYMBOL(iounmap); + +void free_initmem(void) +{ + unsigned long addr; + + addr = (unsigned long)__init_begin; + for (; addr < (unsigned long)__init_end; addr += PAGE_SIZE) { + ClearPageReserved(virt_to_page(addr)); + set_page_count(virt_to_page(addr), 1); + free_page(addr); + totalram_pages++; + } + printk ("Freeing unused kernel memory: %luk freed\n", + ((unsigned long)__init_end - (unsigned long)__init_begin) >> 10); +} + +#ifdef CONFIG_BLK_DEV_INITRD +void free_initrd_mem(unsigned long start, unsigned long end) +{ + if (start < end) + printk ("Freeing initrd memory: %ldk freed\n", (end - start) >> 10); + for (; start < end; start += PAGE_SIZE) { + ClearPageReserved(virt_to_page(start)); + set_page_count(virt_to_page(start), 1); + free_page(start); + totalram_pages++; + } +} +#endif + +static DEFINE_SPINLOCK(mmu_context_lock); +static DEFINE_IDR(mmu_context_idr); + +int init_new_context(struct task_struct *tsk, struct mm_struct *mm) +{ + int index; + int err; + +#ifdef CONFIG_HUGETLB_PAGE + /* We leave htlb_segs as it was, but for a fork, we need to + * clear the huge_pgdir. */ + mm->context.huge_pgdir = NULL; +#endif + +again: + if (!idr_pre_get(&mmu_context_idr, GFP_KERNEL)) + return -ENOMEM; + + spin_lock(&mmu_context_lock); + err = idr_get_new_above(&mmu_context_idr, NULL, 1, &index); + spin_unlock(&mmu_context_lock); + + if (err == -EAGAIN) + goto again; + else if (err) + return err; + + if (index > MAX_CONTEXT) { + idr_remove(&mmu_context_idr, index); + return -ENOMEM; + } + + mm->context.id = index; + + return 0; +} + +void destroy_context(struct mm_struct *mm) +{ + spin_lock(&mmu_context_lock); + idr_remove(&mmu_context_idr, mm->context.id); + spin_unlock(&mmu_context_lock); + + mm->context.id = NO_CONTEXT; + + hugetlb_mm_free_pgd(mm); +} + +/* + * Do very early mm setup. + */ +void __init mm_init_ppc64(void) +{ +#ifndef CONFIG_PPC_ISERIES + unsigned long i; +#endif + + ppc64_boot_msg(0x100, "MM Init"); + + /* This is the story of the IO hole... please, keep seated, + * unfortunately, we are out of oxygen masks at the moment. + * So we need some rough way to tell where your big IO hole + * is. On pmac, it's between 2G and 4G, on POWER3, it's around + * that area as well, on POWER4 we don't have one, etc... + * We need that as a "hint" when sizing the TCE table on POWER3 + * So far, the simplest way that seem work well enough for us it + * to just assume that the first discontinuity in our physical + * RAM layout is the IO hole. That may not be correct in the future + * (and isn't on iSeries but then we don't care ;) + */ + +#ifndef CONFIG_PPC_ISERIES + for (i = 1; i < lmb.memory.cnt; i++) { + unsigned long base, prevbase, prevsize; + + prevbase = lmb.memory.region[i-1].physbase; + prevsize = lmb.memory.region[i-1].size; + base = lmb.memory.region[i].physbase; + if (base > (prevbase + prevsize)) { + io_hole_start = prevbase + prevsize; + io_hole_size = base - (prevbase + prevsize); + break; + } + } +#endif /* CONFIG_PPC_ISERIES */ + if (io_hole_start) + printk("IO Hole assumed to be %lx -> %lx\n", + io_hole_start, io_hole_start + io_hole_size - 1); + + ppc64_boot_msg(0x100, "MM Init Done"); +} + +/* + * This is called by /dev/mem to know if a given address has to + * be mapped non-cacheable or not + */ +int page_is_ram(unsigned long pfn) +{ + int i; + unsigned long paddr = (pfn << PAGE_SHIFT); + + for (i=0; i < lmb.memory.cnt; i++) { + unsigned long base; + +#ifdef CONFIG_MSCHUNKS + base = lmb.memory.region[i].physbase; +#else + base = lmb.memory.region[i].base; +#endif + if ((paddr >= base) && + (paddr < (base + lmb.memory.region[i].size))) { + return 1; + } + } + + return 0; +} +EXPORT_SYMBOL(page_is_ram); + +/* + * Initialize the bootmem system and give it all the memory we + * have available. + */ +#ifndef CONFIG_DISCONTIGMEM +void __init do_init_bootmem(void) +{ + unsigned long i; + unsigned long start, bootmap_pages; + unsigned long total_pages = lmb_end_of_DRAM() >> PAGE_SHIFT; + int boot_mapsize; + + /* + * Find an area to use for the bootmem bitmap. Calculate the size of + * bitmap required as (Total Memory) / PAGE_SIZE / BITS_PER_BYTE. + * Add 1 additional page in case the address isn't page-aligned. + */ + bootmap_pages = bootmem_bootmap_pages(total_pages); + + start = abs_to_phys(lmb_alloc(bootmap_pages<> PAGE_SHIFT, total_pages); + + max_pfn = max_low_pfn; + + /* add all physical memory to the bootmem map. Also find the first */ + for (i=0; i < lmb.memory.cnt; i++) { + unsigned long physbase, size; + + physbase = lmb.memory.region[i].physbase; + size = lmb.memory.region[i].size; + free_bootmem(physbase, size); + } + + /* reserve the sections we're already using */ + for (i=0; i < lmb.reserved.cnt; i++) { + unsigned long physbase = lmb.reserved.region[i].physbase; + unsigned long size = lmb.reserved.region[i].size; + + reserve_bootmem(physbase, size); + } +} + +/* + * paging_init() sets up the page tables - in fact we've already done this. + */ +void __init paging_init(void) +{ + unsigned long zones_size[MAX_NR_ZONES]; + unsigned long zholes_size[MAX_NR_ZONES]; + unsigned long total_ram = lmb_phys_mem_size(); + unsigned long top_of_ram = lmb_end_of_DRAM(); + + printk(KERN_INFO "Top of RAM: 0x%lx, Total RAM: 0x%lx\n", + top_of_ram, total_ram); + printk(KERN_INFO "Memory hole size: %ldMB\n", + (top_of_ram - total_ram) >> 20); + /* + * All pages are DMA-able so we put them all in the DMA zone. + */ + memset(zones_size, 0, sizeof(zones_size)); + memset(zholes_size, 0, sizeof(zholes_size)); + + zones_size[ZONE_DMA] = top_of_ram >> PAGE_SHIFT; + zholes_size[ZONE_DMA] = (top_of_ram - total_ram) >> PAGE_SHIFT; + + free_area_init_node(0, &contig_page_data, zones_size, + __pa(PAGE_OFFSET) >> PAGE_SHIFT, zholes_size); +} +#endif /* CONFIG_DISCONTIGMEM */ + +static struct kcore_list kcore_vmem; + +static int __init setup_kcore(void) +{ + int i; + + for (i=0; i < lmb.memory.cnt; i++) { + unsigned long physbase, size; + struct kcore_list *kcore_mem; + + physbase = lmb.memory.region[i].physbase; + size = lmb.memory.region[i].size; + + /* GFP_ATOMIC to avoid might_sleep warnings during boot */ + kcore_mem = kmalloc(sizeof(struct kcore_list), GFP_ATOMIC); + if (!kcore_mem) + panic("mem_init: kmalloc failed\n"); + + kclist_add(kcore_mem, __va(physbase), size); + } + + kclist_add(&kcore_vmem, (void *)VMALLOC_START, VMALLOC_END-VMALLOC_START); + + return 0; +} +module_init(setup_kcore); + +void __init mem_init(void) +{ +#ifdef CONFIG_DISCONTIGMEM + int nid; +#endif + pg_data_t *pgdat; + unsigned long i; + struct page *page; + unsigned long reservedpages = 0, codesize, initsize, datasize, bsssize; + + num_physpages = max_low_pfn; /* RAM is assumed contiguous */ + high_memory = (void *) __va(max_low_pfn * PAGE_SIZE); + +#ifdef CONFIG_DISCONTIGMEM + for_each_online_node(nid) { + if (NODE_DATA(nid)->node_spanned_pages != 0) { + printk("freeing bootmem node %x\n", nid); + totalram_pages += + free_all_bootmem_node(NODE_DATA(nid)); + } + } +#else + max_mapnr = num_physpages; + totalram_pages += free_all_bootmem(); +#endif + + for_each_pgdat(pgdat) { + for (i = 0; i < pgdat->node_spanned_pages; i++) { + page = pgdat->node_mem_map + i; + if (PageReserved(page)) + reservedpages++; + } + } + + codesize = (unsigned long)&_etext - (unsigned long)&_stext; + initsize = (unsigned long)&__init_end - (unsigned long)&__init_begin; + datasize = (unsigned long)&_edata - (unsigned long)&__init_end; + bsssize = (unsigned long)&__bss_stop - (unsigned long)&__bss_start; + + printk(KERN_INFO "Memory: %luk/%luk available (%luk kernel code, " + "%luk reserved, %luk data, %luk bss, %luk init)\n", + (unsigned long)nr_free_pages() << (PAGE_SHIFT-10), + num_physpages << (PAGE_SHIFT-10), + codesize >> 10, + reservedpages << (PAGE_SHIFT-10), + datasize >> 10, + bsssize >> 10, + initsize >> 10); + + mem_init_done = 1; + +#ifdef CONFIG_PPC_ISERIES + iommu_vio_init(); +#endif + /* Initialize the vDSO */ + vdso_init(); +} + +/* + * This is called when a page has been modified by the kernel. + * It just marks the page as not i-cache clean. We do the i-cache + * flush later when the page is given to a user process, if necessary. + */ +void flush_dcache_page(struct page *page) +{ + if (cpu_has_feature(CPU_FTR_COHERENT_ICACHE)) + return; + /* avoid an atomic op if possible */ + if (test_bit(PG_arch_1, &page->flags)) + clear_bit(PG_arch_1, &page->flags); +} +EXPORT_SYMBOL(flush_dcache_page); + +void clear_user_page(void *page, unsigned long vaddr, struct page *pg) +{ + clear_page(page); + + if (cpu_has_feature(CPU_FTR_COHERENT_ICACHE)) + return; + /* + * We shouldnt have to do this, but some versions of glibc + * require it (ld.so assumes zero filled pages are icache clean) + * - Anton + */ + + /* avoid an atomic op if possible */ + if (test_bit(PG_arch_1, &pg->flags)) + clear_bit(PG_arch_1, &pg->flags); +} +EXPORT_SYMBOL(clear_user_page); + +void copy_user_page(void *vto, void *vfrom, unsigned long vaddr, + struct page *pg) +{ + copy_page(vto, vfrom); + + /* + * We should be able to use the following optimisation, however + * there are two problems. + * Firstly a bug in some versions of binutils meant PLT sections + * were not marked executable. + * Secondly the first word in the GOT section is blrl, used + * to establish the GOT address. Until recently the GOT was + * not marked executable. + * - Anton + */ +#if 0 + if (!vma->vm_file && ((vma->vm_flags & VM_EXEC) == 0)) + return; +#endif + + if (cpu_has_feature(CPU_FTR_COHERENT_ICACHE)) + return; + + /* avoid an atomic op if possible */ + if (test_bit(PG_arch_1, &pg->flags)) + clear_bit(PG_arch_1, &pg->flags); +} + +void flush_icache_user_range(struct vm_area_struct *vma, struct page *page, + unsigned long addr, int len) +{ + unsigned long maddr; + + maddr = (unsigned long)page_address(page) + (addr & ~PAGE_MASK); + flush_icache_range(maddr, maddr + len); +} +EXPORT_SYMBOL(flush_icache_user_range); + +/* + * This is called at the end of handling a user page fault, when the + * fault has been handled by updating a PTE in the linux page tables. + * We use it to preload an HPTE into the hash table corresponding to + * the updated linux PTE. + * + * This must always be called with the mm->page_table_lock held + */ +void update_mmu_cache(struct vm_area_struct *vma, unsigned long ea, + pte_t pte) +{ + unsigned long vsid; + void *pgdir; + pte_t *ptep; + int local = 0; + cpumask_t tmp; + unsigned long flags; + + /* handle i-cache coherency */ + if (!cpu_has_feature(CPU_FTR_COHERENT_ICACHE) && + !cpu_has_feature(CPU_FTR_NOEXECUTE)) { + unsigned long pfn = pte_pfn(pte); + if (pfn_valid(pfn)) { + struct page *page = pfn_to_page(pfn); + if (!PageReserved(page) + && !test_bit(PG_arch_1, &page->flags)) { + __flush_dcache_icache(page_address(page)); + set_bit(PG_arch_1, &page->flags); + } + } + } + + /* We only want HPTEs for linux PTEs that have _PAGE_ACCESSED set */ + if (!pte_young(pte)) + return; + + pgdir = vma->vm_mm->pgd; + if (pgdir == NULL) + return; + + ptep = find_linux_pte(pgdir, ea); + if (!ptep) + return; + + vsid = get_vsid(vma->vm_mm->context.id, ea); + + local_irq_save(flags); + tmp = cpumask_of_cpu(smp_processor_id()); + if (cpus_equal(vma->vm_mm->cpu_vm_mask, tmp)) + local = 1; + + __hash_page(ea, pte_val(pte) & (_PAGE_USER|_PAGE_RW), vsid, ptep, + 0x300, local); + local_irq_restore(flags); +} + +void __iomem * reserve_phb_iospace(unsigned long size) +{ + void __iomem *virt_addr; + + if (phbs_io_bot >= IMALLOC_BASE) + panic("reserve_phb_iospace(): phb io space overflow\n"); + + virt_addr = (void __iomem *) phbs_io_bot; + phbs_io_bot += size; + + return virt_addr; +} + +kmem_cache_t *zero_cache; + +static void zero_ctor(void *pte, kmem_cache_t *cache, unsigned long flags) +{ + memset(pte, 0, PAGE_SIZE); +} + +void pgtable_cache_init(void) +{ + zero_cache = kmem_cache_create("zero", + PAGE_SIZE, + 0, + SLAB_HWCACHE_ALIGN | SLAB_MUST_HWCACHE_ALIGN, + zero_ctor, + NULL); + if (!zero_cache) + panic("pgtable_cache_init(): could not create zero_cache!\n"); +} + +pgprot_t phys_mem_access_prot(struct file *file, unsigned long addr, + unsigned long size, pgprot_t vma_prot) +{ + if (ppc_md.phys_mem_access_prot) + return ppc_md.phys_mem_access_prot(file, addr, size, vma_prot); + + if (!page_is_ram(addr >> PAGE_SHIFT)) + vma_prot = __pgprot(pgprot_val(vma_prot) + | _PAGE_GUARDED | _PAGE_NO_CACHE); + return vma_prot; +} +EXPORT_SYMBOL(phys_mem_access_prot); diff --git a/arch/ppc64/mm/mmap.c b/arch/ppc64/mm/mmap.c new file mode 100644 index 00000000000..fe65f522aff --- /dev/null +++ b/arch/ppc64/mm/mmap.c @@ -0,0 +1,86 @@ +/* + * linux/arch/ppc64/mm/mmap.c + * + * flexible mmap layout support + * + * Copyright 2003-2004 Red Hat Inc., Durham, North Carolina. + * All Rights Reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * + * + * Started by Ingo Molnar + */ + +#include +#include + +/* + * Top of mmap area (just below the process stack). + * + * Leave an at least ~128 MB hole. + */ +#define MIN_GAP (128*1024*1024) +#define MAX_GAP (TASK_SIZE/6*5) + +static inline unsigned long mmap_base(void) +{ + unsigned long gap = current->signal->rlim[RLIMIT_STACK].rlim_cur; + + if (gap < MIN_GAP) + gap = MIN_GAP; + else if (gap > MAX_GAP) + gap = MAX_GAP; + + return TASK_SIZE - (gap & PAGE_MASK); +} + +static inline int mmap_is_legacy(void) +{ + /* + * Force standard allocation for 64 bit programs. + */ + if (!test_thread_flag(TIF_32BIT)) + return 1; + + if (current->personality & ADDR_COMPAT_LAYOUT) + return 1; + + if (current->signal->rlim[RLIMIT_STACK].rlim_cur == RLIM_INFINITY) + return 1; + + return sysctl_legacy_va_layout; +} + +/* + * This function, called very early during the creation of a new + * process VM image, sets up which VM layout function to use: + */ +void arch_pick_mmap_layout(struct mm_struct *mm) +{ + /* + * Fall back to the standard layout if the personality + * bit is set, or if the expected stack growth is unlimited: + */ + if (mmap_is_legacy()) { + mm->mmap_base = TASK_UNMAPPED_BASE; + mm->get_unmapped_area = arch_get_unmapped_area; + mm->unmap_area = arch_unmap_area; + } else { + mm->mmap_base = mmap_base(); + mm->get_unmapped_area = arch_get_unmapped_area_topdown; + mm->unmap_area = arch_unmap_area_topdown; + } +} diff --git a/arch/ppc64/mm/numa.c b/arch/ppc64/mm/numa.c new file mode 100644 index 00000000000..ea862ec643d --- /dev/null +++ b/arch/ppc64/mm/numa.c @@ -0,0 +1,734 @@ +/* + * pSeries NUMA support + * + * Copyright (C) 2002 Anton Blanchard , IBM + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +static int numa_enabled = 1; + +static int numa_debug; +#define dbg(args...) if (numa_debug) { printk(KERN_INFO args); } + +#ifdef DEBUG_NUMA +#define ARRAY_INITIALISER -1 +#else +#define ARRAY_INITIALISER 0 +#endif + +int numa_cpu_lookup_table[NR_CPUS] = { [ 0 ... (NR_CPUS - 1)] = + ARRAY_INITIALISER}; +char *numa_memory_lookup_table; +cpumask_t numa_cpumask_lookup_table[MAX_NUMNODES]; +int nr_cpus_in_node[MAX_NUMNODES] = { [0 ... (MAX_NUMNODES -1)] = 0}; + +struct pglist_data *node_data[MAX_NUMNODES]; +bootmem_data_t __initdata plat_node_bdata[MAX_NUMNODES]; +static int min_common_depth; + +/* + * We need somewhere to store start/span for each node until we have + * allocated the real node_data structures. + */ +static struct { + unsigned long node_start_pfn; + unsigned long node_end_pfn; + unsigned long node_present_pages; +} init_node_data[MAX_NUMNODES] __initdata; + +EXPORT_SYMBOL(node_data); +EXPORT_SYMBOL(numa_cpu_lookup_table); +EXPORT_SYMBOL(numa_memory_lookup_table); +EXPORT_SYMBOL(numa_cpumask_lookup_table); +EXPORT_SYMBOL(nr_cpus_in_node); + +static inline void map_cpu_to_node(int cpu, int node) +{ + numa_cpu_lookup_table[cpu] = node; + if (!(cpu_isset(cpu, numa_cpumask_lookup_table[node]))) { + cpu_set(cpu, numa_cpumask_lookup_table[node]); + nr_cpus_in_node[node]++; + } +} + +#ifdef CONFIG_HOTPLUG_CPU +static void unmap_cpu_from_node(unsigned long cpu) +{ + int node = numa_cpu_lookup_table[cpu]; + + dbg("removing cpu %lu from node %d\n", cpu, node); + + if (cpu_isset(cpu, numa_cpumask_lookup_table[node])) { + cpu_clear(cpu, numa_cpumask_lookup_table[node]); + nr_cpus_in_node[node]--; + } else { + printk(KERN_ERR "WARNING: cpu %lu not found in node %d\n", + cpu, node); + } +} +#endif /* CONFIG_HOTPLUG_CPU */ + +static struct device_node * __devinit find_cpu_node(unsigned int cpu) +{ + unsigned int hw_cpuid = get_hard_smp_processor_id(cpu); + struct device_node *cpu_node = NULL; + unsigned int *interrupt_server, *reg; + int len; + + while ((cpu_node = of_find_node_by_type(cpu_node, "cpu")) != NULL) { + /* Try interrupt server first */ + interrupt_server = (unsigned int *)get_property(cpu_node, + "ibm,ppc-interrupt-server#s", &len); + + len = len / sizeof(u32); + + if (interrupt_server && (len > 0)) { + while (len--) { + if (interrupt_server[len] == hw_cpuid) + return cpu_node; + } + } else { + reg = (unsigned int *)get_property(cpu_node, + "reg", &len); + if (reg && (len > 0) && (reg[0] == hw_cpuid)) + return cpu_node; + } + } + + return NULL; +} + +/* must hold reference to node during call */ +static int *of_get_associativity(struct device_node *dev) +{ + return (unsigned int *)get_property(dev, "ibm,associativity", NULL); +} + +static int of_node_numa_domain(struct device_node *device) +{ + int numa_domain; + unsigned int *tmp; + + if (min_common_depth == -1) + return 0; + + tmp = of_get_associativity(device); + if (tmp && (tmp[0] >= min_common_depth)) { + numa_domain = tmp[min_common_depth]; + } else { + dbg("WARNING: no NUMA information for %s\n", + device->full_name); + numa_domain = 0; + } + return numa_domain; +} + +/* + * In theory, the "ibm,associativity" property may contain multiple + * associativity lists because a resource may be multiply connected + * into the machine. This resource then has different associativity + * characteristics relative to its multiple connections. We ignore + * this for now. We also assume that all cpu and memory sets have + * their distances represented at a common level. This won't be + * true for heirarchical NUMA. + * + * In any case the ibm,associativity-reference-points should give + * the correct depth for a normal NUMA system. + * + * - Dave Hansen + */ +static int __init find_min_common_depth(void) +{ + int depth; + unsigned int *ref_points; + struct device_node *rtas_root; + unsigned int len; + + rtas_root = of_find_node_by_path("/rtas"); + + if (!rtas_root) + return -1; + + /* + * this property is 2 32-bit integers, each representing a level of + * depth in the associativity nodes. The first is for an SMP + * configuration (should be all 0's) and the second is for a normal + * NUMA configuration. + */ + ref_points = (unsigned int *)get_property(rtas_root, + "ibm,associativity-reference-points", &len); + + if ((len >= 1) && ref_points) { + depth = ref_points[1]; + } else { + dbg("WARNING: could not find NUMA " + "associativity reference point\n"); + depth = -1; + } + of_node_put(rtas_root); + + return depth; +} + +static int __init get_mem_addr_cells(void) +{ + struct device_node *memory = NULL; + int rc; + + memory = of_find_node_by_type(memory, "memory"); + if (!memory) + return 0; /* it won't matter */ + + rc = prom_n_addr_cells(memory); + return rc; +} + +static int __init get_mem_size_cells(void) +{ + struct device_node *memory = NULL; + int rc; + + memory = of_find_node_by_type(memory, "memory"); + if (!memory) + return 0; /* it won't matter */ + rc = prom_n_size_cells(memory); + return rc; +} + +static unsigned long read_n_cells(int n, unsigned int **buf) +{ + unsigned long result = 0; + + while (n--) { + result = (result << 32) | **buf; + (*buf)++; + } + return result; +} + +/* + * Figure out to which domain a cpu belongs and stick it there. + * Return the id of the domain used. + */ +static int numa_setup_cpu(unsigned long lcpu) +{ + int numa_domain = 0; + struct device_node *cpu = find_cpu_node(lcpu); + + if (!cpu) { + WARN_ON(1); + goto out; + } + + numa_domain = of_node_numa_domain(cpu); + + if (numa_domain >= num_online_nodes()) { + /* + * POWER4 LPAR uses 0xffff as invalid node, + * dont warn in this case. + */ + if (numa_domain != 0xffff) + printk(KERN_ERR "WARNING: cpu %ld " + "maps to invalid NUMA node %d\n", + lcpu, numa_domain); + numa_domain = 0; + } +out: + node_set_online(numa_domain); + + map_cpu_to_node(lcpu, numa_domain); + + of_node_put(cpu); + + return numa_domain; +} + +static int cpu_numa_callback(struct notifier_block *nfb, + unsigned long action, + void *hcpu) +{ + unsigned long lcpu = (unsigned long)hcpu; + int ret = NOTIFY_DONE; + + switch (action) { + case CPU_UP_PREPARE: + if (min_common_depth == -1 || !numa_enabled) + map_cpu_to_node(lcpu, 0); + else + numa_setup_cpu(lcpu); + ret = NOTIFY_OK; + break; +#ifdef CONFIG_HOTPLUG_CPU + case CPU_DEAD: + case CPU_UP_CANCELED: + unmap_cpu_from_node(lcpu); + break; + ret = NOTIFY_OK; +#endif + } + return ret; +} + +/* + * Check and possibly modify a memory region to enforce the memory limit. + * + * Returns the size the region should have to enforce the memory limit. + * This will either be the original value of size, a truncated value, + * or zero. If the returned value of size is 0 the region should be + * discarded as it lies wholy above the memory limit. + */ +static unsigned long __init numa_enforce_memory_limit(unsigned long start, unsigned long size) +{ + /* + * We use lmb_end_of_DRAM() in here instead of memory_limit because + * we've already adjusted it for the limit and it takes care of + * having memory holes below the limit. + */ + extern unsigned long memory_limit; + + if (! memory_limit) + return size; + + if (start + size <= lmb_end_of_DRAM()) + return size; + + if (start >= lmb_end_of_DRAM()) + return 0; + + return lmb_end_of_DRAM() - start; +} + +static int __init parse_numa_properties(void) +{ + struct device_node *cpu = NULL; + struct device_node *memory = NULL; + int addr_cells, size_cells; + int max_domain = 0; + long entries = lmb_end_of_DRAM() >> MEMORY_INCREMENT_SHIFT; + unsigned long i; + + if (numa_enabled == 0) { + printk(KERN_WARNING "NUMA disabled by user\n"); + return -1; + } + + numa_memory_lookup_table = + (char *)abs_to_virt(lmb_alloc(entries * sizeof(char), 1)); + memset(numa_memory_lookup_table, 0, entries * sizeof(char)); + + for (i = 0; i < entries ; i++) + numa_memory_lookup_table[i] = ARRAY_INITIALISER; + + min_common_depth = find_min_common_depth(); + + dbg("NUMA associativity depth for CPU/Memory: %d\n", min_common_depth); + if (min_common_depth < 0) + return min_common_depth; + + max_domain = numa_setup_cpu(boot_cpuid); + + /* + * Even though we connect cpus to numa domains later in SMP init, + * we need to know the maximum node id now. This is because each + * node id must have NODE_DATA etc backing it. + * As a result of hotplug we could still have cpus appear later on + * with larger node ids. In that case we force the cpu into node 0. + */ + for_each_cpu(i) { + int numa_domain; + + cpu = find_cpu_node(i); + + if (cpu) { + numa_domain = of_node_numa_domain(cpu); + of_node_put(cpu); + + if (numa_domain < MAX_NUMNODES && + max_domain < numa_domain) + max_domain = numa_domain; + } + } + + addr_cells = get_mem_addr_cells(); + size_cells = get_mem_size_cells(); + memory = NULL; + while ((memory = of_find_node_by_type(memory, "memory")) != NULL) { + unsigned long start; + unsigned long size; + int numa_domain; + int ranges; + unsigned int *memcell_buf; + unsigned int len; + + memcell_buf = (unsigned int *)get_property(memory, "reg", &len); + if (!memcell_buf || len <= 0) + continue; + + ranges = memory->n_addrs; +new_range: + /* these are order-sensitive, and modify the buffer pointer */ + start = read_n_cells(addr_cells, &memcell_buf); + size = read_n_cells(size_cells, &memcell_buf); + + start = _ALIGN_DOWN(start, MEMORY_INCREMENT); + size = _ALIGN_UP(size, MEMORY_INCREMENT); + + numa_domain = of_node_numa_domain(memory); + + if (numa_domain >= MAX_NUMNODES) { + if (numa_domain != 0xffff) + printk(KERN_ERR "WARNING: memory at %lx maps " + "to invalid NUMA node %d\n", start, + numa_domain); + numa_domain = 0; + } + + if (max_domain < numa_domain) + max_domain = numa_domain; + + if (! (size = numa_enforce_memory_limit(start, size))) { + if (--ranges) + goto new_range; + else + continue; + } + + /* + * Initialize new node struct, or add to an existing one. + */ + if (init_node_data[numa_domain].node_end_pfn) { + if ((start / PAGE_SIZE) < + init_node_data[numa_domain].node_start_pfn) + init_node_data[numa_domain].node_start_pfn = + start / PAGE_SIZE; + if (((start / PAGE_SIZE) + (size / PAGE_SIZE)) > + init_node_data[numa_domain].node_end_pfn) + init_node_data[numa_domain].node_end_pfn = + (start / PAGE_SIZE) + + (size / PAGE_SIZE); + + init_node_data[numa_domain].node_present_pages += + size / PAGE_SIZE; + } else { + node_set_online(numa_domain); + + init_node_data[numa_domain].node_start_pfn = + start / PAGE_SIZE; + init_node_data[numa_domain].node_end_pfn = + init_node_data[numa_domain].node_start_pfn + + size / PAGE_SIZE; + init_node_data[numa_domain].node_present_pages = + size / PAGE_SIZE; + } + + for (i = start ; i < (start+size); i += MEMORY_INCREMENT) + numa_memory_lookup_table[i >> MEMORY_INCREMENT_SHIFT] = + numa_domain; + + if (--ranges) + goto new_range; + } + + for (i = 0; i <= max_domain; i++) + node_set_online(i); + + return 0; +} + +static void __init setup_nonnuma(void) +{ + unsigned long top_of_ram = lmb_end_of_DRAM(); + unsigned long total_ram = lmb_phys_mem_size(); + unsigned long i; + + printk(KERN_INFO "Top of RAM: 0x%lx, Total RAM: 0x%lx\n", + top_of_ram, total_ram); + printk(KERN_INFO "Memory hole size: %ldMB\n", + (top_of_ram - total_ram) >> 20); + + if (!numa_memory_lookup_table) { + long entries = top_of_ram >> MEMORY_INCREMENT_SHIFT; + numa_memory_lookup_table = + (char *)abs_to_virt(lmb_alloc(entries * sizeof(char), 1)); + memset(numa_memory_lookup_table, 0, entries * sizeof(char)); + for (i = 0; i < entries ; i++) + numa_memory_lookup_table[i] = ARRAY_INITIALISER; + } + + map_cpu_to_node(boot_cpuid, 0); + + node_set_online(0); + + init_node_data[0].node_start_pfn = 0; + init_node_data[0].node_end_pfn = lmb_end_of_DRAM() / PAGE_SIZE; + init_node_data[0].node_present_pages = total_ram / PAGE_SIZE; + + for (i = 0 ; i < top_of_ram; i += MEMORY_INCREMENT) + numa_memory_lookup_table[i >> MEMORY_INCREMENT_SHIFT] = 0; +} + +static void __init dump_numa_topology(void) +{ + unsigned int node; + unsigned int count; + + if (min_common_depth == -1 || !numa_enabled) + return; + + for_each_online_node(node) { + unsigned long i; + + printk(KERN_INFO "Node %d Memory:", node); + + count = 0; + + for (i = 0; i < lmb_end_of_DRAM(); i += MEMORY_INCREMENT) { + if (numa_memory_lookup_table[i >> MEMORY_INCREMENT_SHIFT] == node) { + if (count == 0) + printk(" 0x%lx", i); + ++count; + } else { + if (count > 0) + printk("-0x%lx", i); + count = 0; + } + } + + if (count > 0) + printk("-0x%lx", i); + printk("\n"); + } + return; +} + +/* + * Allocate some memory, satisfying the lmb or bootmem allocator where + * required. nid is the preferred node and end is the physical address of + * the highest address in the node. + * + * Returns the physical address of the memory. + */ +static unsigned long careful_allocation(int nid, unsigned long size, + unsigned long align, unsigned long end) +{ + unsigned long ret = lmb_alloc_base(size, align, end); + + /* retry over all memory */ + if (!ret) + ret = lmb_alloc_base(size, align, lmb_end_of_DRAM()); + + if (!ret) + panic("numa.c: cannot allocate %lu bytes on node %d", + size, nid); + + /* + * If the memory came from a previously allocated node, we must + * retry with the bootmem allocator. + */ + if (pa_to_nid(ret) < nid) { + nid = pa_to_nid(ret); + ret = (unsigned long)__alloc_bootmem_node(NODE_DATA(nid), + size, align, 0); + + if (!ret) + panic("numa.c: cannot allocate %lu bytes on node %d", + size, nid); + + ret = virt_to_abs(ret); + + dbg("alloc_bootmem %lx %lx\n", ret, size); + } + + return ret; +} + +void __init do_init_bootmem(void) +{ + int nid; + int addr_cells, size_cells; + struct device_node *memory = NULL; + static struct notifier_block ppc64_numa_nb = { + .notifier_call = cpu_numa_callback, + .priority = 1 /* Must run before sched domains notifier. */ + }; + + min_low_pfn = 0; + max_low_pfn = lmb_end_of_DRAM() >> PAGE_SHIFT; + max_pfn = max_low_pfn; + + if (parse_numa_properties()) + setup_nonnuma(); + else + dump_numa_topology(); + + register_cpu_notifier(&ppc64_numa_nb); + + for_each_online_node(nid) { + unsigned long start_paddr, end_paddr; + int i; + unsigned long bootmem_paddr; + unsigned long bootmap_pages; + + start_paddr = init_node_data[nid].node_start_pfn * PAGE_SIZE; + end_paddr = init_node_data[nid].node_end_pfn * PAGE_SIZE; + + /* Allocate the node structure node local if possible */ + NODE_DATA(nid) = (struct pglist_data *)careful_allocation(nid, + sizeof(struct pglist_data), + SMP_CACHE_BYTES, end_paddr); + NODE_DATA(nid) = abs_to_virt(NODE_DATA(nid)); + memset(NODE_DATA(nid), 0, sizeof(struct pglist_data)); + + dbg("node %d\n", nid); + dbg("NODE_DATA() = %p\n", NODE_DATA(nid)); + + NODE_DATA(nid)->bdata = &plat_node_bdata[nid]; + NODE_DATA(nid)->node_start_pfn = + init_node_data[nid].node_start_pfn; + NODE_DATA(nid)->node_spanned_pages = + end_paddr - start_paddr; + + if (NODE_DATA(nid)->node_spanned_pages == 0) + continue; + + dbg("start_paddr = %lx\n", start_paddr); + dbg("end_paddr = %lx\n", end_paddr); + + bootmap_pages = bootmem_bootmap_pages((end_paddr - start_paddr) >> PAGE_SHIFT); + + bootmem_paddr = careful_allocation(nid, + bootmap_pages << PAGE_SHIFT, + PAGE_SIZE, end_paddr); + memset(abs_to_virt(bootmem_paddr), 0, + bootmap_pages << PAGE_SHIFT); + dbg("bootmap_paddr = %lx\n", bootmem_paddr); + + init_bootmem_node(NODE_DATA(nid), bootmem_paddr >> PAGE_SHIFT, + start_paddr >> PAGE_SHIFT, + end_paddr >> PAGE_SHIFT); + + /* + * We need to do another scan of all memory sections to + * associate memory with the correct node. + */ + addr_cells = get_mem_addr_cells(); + size_cells = get_mem_size_cells(); + memory = NULL; + while ((memory = of_find_node_by_type(memory, "memory")) != NULL) { + unsigned long mem_start, mem_size; + int numa_domain, ranges; + unsigned int *memcell_buf; + unsigned int len; + + memcell_buf = (unsigned int *)get_property(memory, "reg", &len); + if (!memcell_buf || len <= 0) + continue; + + ranges = memory->n_addrs; /* ranges in cell */ +new_range: + mem_start = read_n_cells(addr_cells, &memcell_buf); + mem_size = read_n_cells(size_cells, &memcell_buf); + numa_domain = numa_enabled ? of_node_numa_domain(memory) : 0; + + if (numa_domain != nid) + continue; + + mem_size = numa_enforce_memory_limit(mem_start, mem_size); + if (mem_size) { + dbg("free_bootmem %lx %lx\n", mem_start, mem_size); + free_bootmem_node(NODE_DATA(nid), mem_start, mem_size); + } + + if (--ranges) /* process all ranges in cell */ + goto new_range; + } + + /* + * Mark reserved regions on this node + */ + for (i = 0; i < lmb.reserved.cnt; i++) { + unsigned long physbase = lmb.reserved.region[i].physbase; + unsigned long size = lmb.reserved.region[i].size; + + if (pa_to_nid(physbase) != nid && + pa_to_nid(physbase+size-1) != nid) + continue; + + if (physbase < end_paddr && + (physbase+size) > start_paddr) { + /* overlaps */ + if (physbase < start_paddr) { + size -= start_paddr - physbase; + physbase = start_paddr; + } + + if (size > end_paddr - physbase) + size = end_paddr - physbase; + + dbg("reserve_bootmem %lx %lx\n", physbase, + size); + reserve_bootmem_node(NODE_DATA(nid), physbase, + size); + } + } + } +} + +void __init paging_init(void) +{ + unsigned long zones_size[MAX_NR_ZONES]; + unsigned long zholes_size[MAX_NR_ZONES]; + int nid; + + memset(zones_size, 0, sizeof(zones_size)); + memset(zholes_size, 0, sizeof(zholes_size)); + + for_each_online_node(nid) { + unsigned long start_pfn; + unsigned long end_pfn; + + start_pfn = init_node_data[nid].node_start_pfn; + end_pfn = init_node_data[nid].node_end_pfn; + + zones_size[ZONE_DMA] = end_pfn - start_pfn; + zholes_size[ZONE_DMA] = zones_size[ZONE_DMA] - + init_node_data[nid].node_present_pages; + + dbg("free_area_init node %d %lx %lx (hole: %lx)\n", nid, + zones_size[ZONE_DMA], start_pfn, zholes_size[ZONE_DMA]); + + free_area_init_node(nid, NODE_DATA(nid), zones_size, + start_pfn, zholes_size); + } +} + +static int __init early_numa(char *p) +{ + if (!p) + return 0; + + if (strstr(p, "off")) + numa_enabled = 0; + + if (strstr(p, "debug")) + numa_debug = 1; + + return 0; +} +early_param("numa", early_numa); diff --git a/arch/ppc64/mm/slb.c b/arch/ppc64/mm/slb.c new file mode 100644 index 00000000000..6a20773f695 --- /dev/null +++ b/arch/ppc64/mm/slb.c @@ -0,0 +1,159 @@ +/* + * PowerPC64 SLB support. + * + * Copyright (C) 2004 David Gibson , IBM + * Based on earlier code writteh by: + * Dave Engebretsen and Mike Corrigan {engebret|mikejc}@us.ibm.com + * Copyright (c) 2001 Dave Engebretsen + * Copyright (C) 2002 Anton Blanchard , IBM + * + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +#include +#include +#include +#include +#include +#include + +extern void slb_allocate(unsigned long ea); + +static inline unsigned long mk_esid_data(unsigned long ea, unsigned long slot) +{ + return (ea & ESID_MASK) | SLB_ESID_V | slot; +} + +static inline unsigned long mk_vsid_data(unsigned long ea, unsigned long flags) +{ + return (get_kernel_vsid(ea) << SLB_VSID_SHIFT) | flags; +} + +static inline void create_slbe(unsigned long ea, unsigned long vsid, + unsigned long flags, unsigned long entry) +{ + asm volatile("slbmte %0,%1" : + : "r" (mk_vsid_data(ea, flags)), + "r" (mk_esid_data(ea, entry)) + : "memory" ); +} + +static void slb_flush_and_rebolt(void) +{ + /* If you change this make sure you change SLB_NUM_BOLTED + * appropriately too. */ + unsigned long ksp_flags = SLB_VSID_KERNEL; + unsigned long ksp_esid_data; + + WARN_ON(!irqs_disabled()); + + if (cpu_has_feature(CPU_FTR_16M_PAGE)) + ksp_flags |= SLB_VSID_L; + + ksp_esid_data = mk_esid_data(get_paca()->kstack, 2); + if ((ksp_esid_data & ESID_MASK) == KERNELBASE) + ksp_esid_data &= ~SLB_ESID_V; + + /* We need to do this all in asm, so we're sure we don't touch + * the stack between the slbia and rebolting it. */ + asm volatile("isync\n" + "slbia\n" + /* Slot 1 - first VMALLOC segment */ + "slbmte %0,%1\n" + /* Slot 2 - kernel stack */ + "slbmte %2,%3\n" + "isync" + :: "r"(mk_vsid_data(VMALLOCBASE, SLB_VSID_KERNEL)), + "r"(mk_esid_data(VMALLOCBASE, 1)), + "r"(mk_vsid_data(ksp_esid_data, ksp_flags)), + "r"(ksp_esid_data) + : "memory"); +} + +/* Flush all user entries from the segment table of the current processor. */ +void switch_slb(struct task_struct *tsk, struct mm_struct *mm) +{ + unsigned long offset = get_paca()->slb_cache_ptr; + unsigned long esid_data = 0; + unsigned long pc = KSTK_EIP(tsk); + unsigned long stack = KSTK_ESP(tsk); + unsigned long unmapped_base; + + if (offset <= SLB_CACHE_ENTRIES) { + int i; + asm volatile("isync" : : : "memory"); + for (i = 0; i < offset; i++) { + esid_data = (unsigned long)get_paca()->slb_cache[i] + << SID_SHIFT; + asm volatile("slbie %0" : : "r" (esid_data)); + } + asm volatile("isync" : : : "memory"); + } else { + slb_flush_and_rebolt(); + } + + /* Workaround POWER5 < DD2.1 issue */ + if (offset == 1 || offset > SLB_CACHE_ENTRIES) + asm volatile("slbie %0" : : "r" (esid_data)); + + get_paca()->slb_cache_ptr = 0; + get_paca()->context = mm->context; + + /* + * preload some userspace segments into the SLB. + */ + if (test_tsk_thread_flag(tsk, TIF_32BIT)) + unmapped_base = TASK_UNMAPPED_BASE_USER32; + else + unmapped_base = TASK_UNMAPPED_BASE_USER64; + + if (pc >= KERNELBASE) + return; + slb_allocate(pc); + + if (GET_ESID(pc) == GET_ESID(stack)) + return; + + if (stack >= KERNELBASE) + return; + slb_allocate(stack); + + if ((GET_ESID(pc) == GET_ESID(unmapped_base)) + || (GET_ESID(stack) == GET_ESID(unmapped_base))) + return; + + if (unmapped_base >= KERNELBASE) + return; + slb_allocate(unmapped_base); +} + +void slb_initialize(void) +{ + /* On iSeries the bolted entries have already been set up by + * the hypervisor from the lparMap data in head.S */ +#ifndef CONFIG_PPC_ISERIES + unsigned long flags = SLB_VSID_KERNEL; + + /* Invalidate the entire SLB (even slot 0) & all the ERATS */ + if (cpu_has_feature(CPU_FTR_16M_PAGE)) + flags |= SLB_VSID_L; + + asm volatile("isync":::"memory"); + asm volatile("slbmte %0,%0"::"r" (0) : "memory"); + asm volatile("isync; slbia; isync":::"memory"); + create_slbe(KERNELBASE, get_kernel_vsid(KERNELBASE), flags, 0); + create_slbe(VMALLOCBASE, get_kernel_vsid(KERNELBASE), + SLB_VSID_KERNEL, 1); + /* We don't bolt the stack for the time being - we're in boot, + * so the stack is in the bolted segment. By the time it goes + * elsewhere, we'll call _switch() which will bolt in the new + * one. */ + asm volatile("isync":::"memory"); +#endif + + get_paca()->stab_rr = SLB_NUM_BOLTED; +} diff --git a/arch/ppc64/mm/slb_low.S b/arch/ppc64/mm/slb_low.S new file mode 100644 index 00000000000..8379d678f70 --- /dev/null +++ b/arch/ppc64/mm/slb_low.S @@ -0,0 +1,154 @@ +/* + * arch/ppc64/mm/slb_low.S + * + * Low-level SLB routines + * + * Copyright (C) 2004 David Gibson , IBM + * + * Based on earlier C version: + * Dave Engebretsen and Mike Corrigan {engebret|mikejc}@us.ibm.com + * Copyright (c) 2001 Dave Engebretsen + * Copyright (C) 2002 Anton Blanchard , IBM + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +#include +#include +#include +#include +#include +#include +#include + +/* void slb_allocate(unsigned long ea); + * + * Create an SLB entry for the given EA (user or kernel). + * r3 = faulting address, r13 = PACA + * r9, r10, r11 are clobbered by this function + * No other registers are examined or changed. + */ +_GLOBAL(slb_allocate) + /* + * First find a slot, round robin. Previously we tried to find + * a free slot first but that took too long. Unfortunately we + * dont have any LRU information to help us choose a slot. + */ +#ifdef CONFIG_PPC_ISERIES + /* + * On iSeries, the "bolted" stack segment can be cast out on + * shared processor switch so we need to check for a miss on + * it and restore it to the right slot. + */ + ld r9,PACAKSAVE(r13) + clrrdi r9,r9,28 + clrrdi r11,r3,28 + li r10,SLB_NUM_BOLTED-1 /* Stack goes in last bolted slot */ + cmpld r9,r11 + beq 3f +#endif /* CONFIG_PPC_ISERIES */ + + ld r10,PACASTABRR(r13) + addi r10,r10,1 + /* use a cpu feature mask if we ever change our slb size */ + cmpldi r10,SLB_NUM_ENTRIES + + blt+ 4f + li r10,SLB_NUM_BOLTED + +4: + std r10,PACASTABRR(r13) +3: + /* r3 = faulting address, r10 = entry */ + + srdi r9,r3,60 /* get region */ + srdi r3,r3,28 /* get esid */ + cmpldi cr7,r9,0xc /* cmp KERNELBASE for later use */ + + rldimi r10,r3,28,0 /* r10= ESID<<28 | entry */ + oris r10,r10,SLB_ESID_V@h /* r10 |= SLB_ESID_V */ + + /* r3 = esid, r10 = esid_data, cr7 = <>KERNELBASE */ + + blt cr7,0f /* user or kernel? */ + + /* kernel address: proto-VSID = ESID */ + /* WARNING - MAGIC: we don't use the VSID 0xfffffffff, but + * this code will generate the protoVSID 0xfffffffff for the + * top segment. That's ok, the scramble below will translate + * it to VSID 0, which is reserved as a bad VSID - one which + * will never have any pages in it. */ + li r11,SLB_VSID_KERNEL +BEGIN_FTR_SECTION + bne cr7,9f + li r11,(SLB_VSID_KERNEL|SLB_VSID_L) +END_FTR_SECTION_IFSET(CPU_FTR_16M_PAGE) + b 9f + +0: /* user address: proto-VSID = context<<15 | ESID */ + li r11,SLB_VSID_USER + + srdi. r9,r3,13 + bne- 8f /* invalid ea bits set */ + +#ifdef CONFIG_HUGETLB_PAGE +BEGIN_FTR_SECTION + /* check against the hugepage ranges */ + cmpldi r3,(TASK_HPAGE_END>>SID_SHIFT) + bge 6f /* >= TASK_HPAGE_END */ + cmpldi r3,(TASK_HPAGE_BASE>>SID_SHIFT) + bge 5f /* TASK_HPAGE_BASE..TASK_HPAGE_END */ + cmpldi r3,16 + bge 6f /* 4GB..TASK_HPAGE_BASE */ + + lhz r9,PACAHTLBSEGS(r13) + srd r9,r9,r3 + andi. r9,r9,1 + beq 6f + +5: /* this is a hugepage user address */ + li r11,(SLB_VSID_USER|SLB_VSID_L) +END_FTR_SECTION_IFSET(CPU_FTR_16M_PAGE) +#endif /* CONFIG_HUGETLB_PAGE */ + +6: ld r9,PACACONTEXTID(r13) + rldimi r3,r9,USER_ESID_BITS,0 + +9: /* r3 = protovsid, r11 = flags, r10 = esid_data, cr7 = <>KERNELBASE */ + ASM_VSID_SCRAMBLE(r3,r9) + + rldimi r11,r3,SLB_VSID_SHIFT,16 /* combine VSID and flags */ + + /* + * No need for an isync before or after this slbmte. The exception + * we enter with and the rfid we exit with are context synchronizing. + */ + slbmte r11,r10 + + bgelr cr7 /* we're done for kernel addresses */ + + /* Update the slb cache */ + lhz r3,PACASLBCACHEPTR(r13) /* offset = paca->slb_cache_ptr */ + cmpldi r3,SLB_CACHE_ENTRIES + bge 1f + + /* still room in the slb cache */ + sldi r11,r3,1 /* r11 = offset * sizeof(u16) */ + rldicl r10,r10,36,28 /* get low 16 bits of the ESID */ + add r11,r11,r13 /* r11 = (u16 *)paca + offset */ + sth r10,PACASLBCACHE(r11) /* paca->slb_cache[offset] = esid */ + addi r3,r3,1 /* offset++ */ + b 2f +1: /* offset >= SLB_CACHE_ENTRIES */ + li r3,SLB_CACHE_ENTRIES+1 +2: + sth r3,PACASLBCACHEPTR(r13) /* paca->slb_cache_ptr = offset */ + blr + +8: /* invalid EA */ + li r3,0 /* BAD_VSID */ + li r11,SLB_VSID_USER /* flags don't much matter */ + b 9b diff --git a/arch/ppc64/mm/stab.c b/arch/ppc64/mm/stab.c new file mode 100644 index 00000000000..31491131d5e --- /dev/null +++ b/arch/ppc64/mm/stab.c @@ -0,0 +1,239 @@ +/* + * PowerPC64 Segment Translation Support. + * + * Dave Engebretsen and Mike Corrigan {engebret|mikejc}@us.ibm.com + * Copyright (c) 2001 Dave Engebretsen + * + * Copyright (C) 2002 Anton Blanchard , IBM + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ + +#include +#include +#include +#include +#include +#include + +/* Both the segment table and SLB code uses the following cache */ +#define NR_STAB_CACHE_ENTRIES 8 +DEFINE_PER_CPU(long, stab_cache_ptr); +DEFINE_PER_CPU(long, stab_cache[NR_STAB_CACHE_ENTRIES]); + +/* + * Create a segment table entry for the given esid/vsid pair. + */ +static int make_ste(unsigned long stab, unsigned long esid, unsigned long vsid) +{ + unsigned long esid_data, vsid_data; + unsigned long entry, group, old_esid, castout_entry, i; + unsigned int global_entry; + struct stab_entry *ste, *castout_ste; + unsigned long kernel_segment = (esid << SID_SHIFT) >= KERNELBASE; + + vsid_data = vsid << STE_VSID_SHIFT; + esid_data = esid << SID_SHIFT | STE_ESID_KP | STE_ESID_V; + if (! kernel_segment) + esid_data |= STE_ESID_KS; + + /* Search the primary group first. */ + global_entry = (esid & 0x1f) << 3; + ste = (struct stab_entry *)(stab | ((esid & 0x1f) << 7)); + + /* Find an empty entry, if one exists. */ + for (group = 0; group < 2; group++) { + for (entry = 0; entry < 8; entry++, ste++) { + if (!(ste->esid_data & STE_ESID_V)) { + ste->vsid_data = vsid_data; + asm volatile("eieio":::"memory"); + ste->esid_data = esid_data; + return (global_entry | entry); + } + } + /* Now search the secondary group. */ + global_entry = ((~esid) & 0x1f) << 3; + ste = (struct stab_entry *)(stab | (((~esid) & 0x1f) << 7)); + } + + /* + * Could not find empty entry, pick one with a round robin selection. + * Search all entries in the two groups. + */ + castout_entry = get_paca()->stab_rr; + for (i = 0; i < 16; i++) { + if (castout_entry < 8) { + global_entry = (esid & 0x1f) << 3; + ste = (struct stab_entry *)(stab | ((esid & 0x1f) << 7)); + castout_ste = ste + castout_entry; + } else { + global_entry = ((~esid) & 0x1f) << 3; + ste = (struct stab_entry *)(stab | (((~esid) & 0x1f) << 7)); + castout_ste = ste + (castout_entry - 8); + } + + /* Dont cast out the first kernel segment */ + if ((castout_ste->esid_data & ESID_MASK) != KERNELBASE) + break; + + castout_entry = (castout_entry + 1) & 0xf; + } + + get_paca()->stab_rr = (castout_entry + 1) & 0xf; + + /* Modify the old entry to the new value. */ + + /* Force previous translations to complete. DRENG */ + asm volatile("isync" : : : "memory"); + + old_esid = castout_ste->esid_data >> SID_SHIFT; + castout_ste->esid_data = 0; /* Invalidate old entry */ + + asm volatile("sync" : : : "memory"); /* Order update */ + + castout_ste->vsid_data = vsid_data; + asm volatile("eieio" : : : "memory"); /* Order update */ + castout_ste->esid_data = esid_data; + + asm volatile("slbie %0" : : "r" (old_esid << SID_SHIFT)); + /* Ensure completion of slbie */ + asm volatile("sync" : : : "memory"); + + return (global_entry | (castout_entry & 0x7)); +} + +/* + * Allocate a segment table entry for the given ea and mm + */ +static int __ste_allocate(unsigned long ea, struct mm_struct *mm) +{ + unsigned long vsid; + unsigned char stab_entry; + unsigned long offset; + + /* Kernel or user address? */ + if (ea >= KERNELBASE) { + vsid = get_kernel_vsid(ea); + } else { + if ((ea >= TASK_SIZE_USER64) || (! mm)) + return 1; + + vsid = get_vsid(mm->context.id, ea); + } + + stab_entry = make_ste(get_paca()->stab_addr, GET_ESID(ea), vsid); + + if (ea < KERNELBASE) { + offset = __get_cpu_var(stab_cache_ptr); + if (offset < NR_STAB_CACHE_ENTRIES) + __get_cpu_var(stab_cache[offset++]) = stab_entry; + else + offset = NR_STAB_CACHE_ENTRIES+1; + __get_cpu_var(stab_cache_ptr) = offset; + + /* Order update */ + asm volatile("sync":::"memory"); + } + + return 0; +} + +int ste_allocate(unsigned long ea) +{ + return __ste_allocate(ea, current->mm); +} + +/* + * Do the segment table work for a context switch: flush all user + * entries from the table, then preload some probably useful entries + * for the new task + */ +void switch_stab(struct task_struct *tsk, struct mm_struct *mm) +{ + struct stab_entry *stab = (struct stab_entry *) get_paca()->stab_addr; + struct stab_entry *ste; + unsigned long offset = __get_cpu_var(stab_cache_ptr); + unsigned long pc = KSTK_EIP(tsk); + unsigned long stack = KSTK_ESP(tsk); + unsigned long unmapped_base; + + /* Force previous translations to complete. DRENG */ + asm volatile("isync" : : : "memory"); + + if (offset <= NR_STAB_CACHE_ENTRIES) { + int i; + + for (i = 0; i < offset; i++) { + ste = stab + __get_cpu_var(stab_cache[i]); + ste->esid_data = 0; /* invalidate entry */ + } + } else { + unsigned long entry; + + /* Invalidate all entries. */ + ste = stab; + + /* Never flush the first entry. */ + ste += 1; + for (entry = 1; + entry < (PAGE_SIZE / sizeof(struct stab_entry)); + entry++, ste++) { + unsigned long ea; + ea = ste->esid_data & ESID_MASK; + if (ea < KERNELBASE) { + ste->esid_data = 0; + } + } + } + + asm volatile("sync; slbia; sync":::"memory"); + + __get_cpu_var(stab_cache_ptr) = 0; + + /* Now preload some entries for the new task */ + if (test_tsk_thread_flag(tsk, TIF_32BIT)) + unmapped_base = TASK_UNMAPPED_BASE_USER32; + else + unmapped_base = TASK_UNMAPPED_BASE_USER64; + + __ste_allocate(pc, mm); + + if (GET_ESID(pc) == GET_ESID(stack)) + return; + + __ste_allocate(stack, mm); + + if ((GET_ESID(pc) == GET_ESID(unmapped_base)) + || (GET_ESID(stack) == GET_ESID(unmapped_base))) + return; + + __ste_allocate(unmapped_base, mm); + + /* Order update */ + asm volatile("sync" : : : "memory"); +} + +extern void slb_initialize(void); + +/* + * Build an entry for the base kernel segment and put it into + * the segment table or SLB. All other segment table or SLB + * entries are faulted in. + */ +void stab_initialize(unsigned long stab) +{ + unsigned long vsid = get_kernel_vsid(KERNELBASE); + + if (cpu_has_feature(CPU_FTR_SLB)) { + slb_initialize(); + } else { + asm volatile("isync; slbia; isync":::"memory"); + make_ste(stab, GET_ESID(KERNELBASE), vsid); + + /* Order update */ + asm volatile("sync":::"memory"); + } +} diff --git a/arch/ppc64/mm/tlb.c b/arch/ppc64/mm/tlb.c new file mode 100644 index 00000000000..26f0172c452 --- /dev/null +++ b/arch/ppc64/mm/tlb.c @@ -0,0 +1,180 @@ +/* + * This file contains the routines for flushing entries from the + * TLB and MMU hash table. + * + * Derived from arch/ppc64/mm/init.c: + * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org) + * + * Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au) + * and Cort Dougan (PReP) (cort@cs.nmt.edu) + * Copyright (C) 1996 Paul Mackerras + * Amiga/APUS changes by Jesper Skov (jskov@cygnus.co.uk). + * + * Derived from "arch/i386/mm/init.c" + * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds + * + * Dave Engebretsen + * Rework for PPC64 port. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +DEFINE_PER_CPU(struct ppc64_tlb_batch, ppc64_tlb_batch); + +/* This is declared as we are using the more or less generic + * include/asm-ppc64/tlb.h file -- tgall + */ +DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); +DEFINE_PER_CPU(struct pte_freelist_batch *, pte_freelist_cur); +unsigned long pte_freelist_forced_free; + +void __pte_free_tlb(struct mmu_gather *tlb, struct page *ptepage) +{ + /* This is safe as we are holding page_table_lock */ + cpumask_t local_cpumask = cpumask_of_cpu(smp_processor_id()); + struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur); + + if (atomic_read(&tlb->mm->mm_users) < 2 || + cpus_equal(tlb->mm->cpu_vm_mask, local_cpumask)) { + pte_free(ptepage); + return; + } + + if (*batchp == NULL) { + *batchp = (struct pte_freelist_batch *)__get_free_page(GFP_ATOMIC); + if (*batchp == NULL) { + pte_free_now(ptepage); + return; + } + (*batchp)->index = 0; + } + (*batchp)->pages[(*batchp)->index++] = ptepage; + if ((*batchp)->index == PTE_FREELIST_SIZE) { + pte_free_submit(*batchp); + *batchp = NULL; + } +} + +/* + * Update the MMU hash table to correspond with a change to + * a Linux PTE. If wrprot is true, it is permissible to + * change the existing HPTE to read-only rather than removing it + * (if we remove it we should clear the _PTE_HPTEFLAGS bits). + */ +void hpte_update(struct mm_struct *mm, unsigned long addr, + unsigned long pte, int wrprot) +{ + int i; + unsigned long context = 0; + struct ppc64_tlb_batch *batch = &__get_cpu_var(ppc64_tlb_batch); + + if (REGION_ID(addr) == USER_REGION_ID) + context = mm->context.id; + i = batch->index; + + /* + * This can happen when we are in the middle of a TLB batch and + * we encounter memory pressure (eg copy_page_range when it tries + * to allocate a new pte). If we have to reclaim memory and end + * up scanning and resetting referenced bits then our batch context + * will change mid stream. + */ + if (unlikely(i != 0 && context != batch->context)) { + flush_tlb_pending(); + i = 0; + } + + if (i == 0) { + batch->context = context; + batch->mm = mm; + } + batch->pte[i] = __pte(pte); + batch->addr[i] = addr; + batch->index = ++i; + if (i >= PPC64_TLB_BATCH_NR) + flush_tlb_pending(); +} + +void __flush_tlb_pending(struct ppc64_tlb_batch *batch) +{ + int i; + int cpu; + cpumask_t tmp; + int local = 0; + + BUG_ON(in_interrupt()); + + cpu = get_cpu(); + i = batch->index; + tmp = cpumask_of_cpu(cpu); + if (cpus_equal(batch->mm->cpu_vm_mask, tmp)) + local = 1; + + if (i == 1) + flush_hash_page(batch->context, batch->addr[0], batch->pte[0], + local); + else + flush_hash_range(batch->context, i, local); + batch->index = 0; + put_cpu(); +} + +#ifdef CONFIG_SMP +static void pte_free_smp_sync(void *arg) +{ + /* Do nothing, just ensure we sync with all CPUs */ +} +#endif + +/* This is only called when we are critically out of memory + * (and fail to get a page in pte_free_tlb). + */ +void pte_free_now(struct page *ptepage) +{ + pte_freelist_forced_free++; + + smp_call_function(pte_free_smp_sync, NULL, 0, 1); + + pte_free(ptepage); +} + +static void pte_free_rcu_callback(struct rcu_head *head) +{ + struct pte_freelist_batch *batch = + container_of(head, struct pte_freelist_batch, rcu); + unsigned int i; + + for (i = 0; i < batch->index; i++) + pte_free(batch->pages[i]); + free_page((unsigned long)batch); +} + +void pte_free_submit(struct pte_freelist_batch *batch) +{ + INIT_RCU_HEAD(&batch->rcu); + call_rcu(&batch->rcu, pte_free_rcu_callback); +} + +void pte_free_finish(void) +{ + /* This is safe as we are holding page_table_lock */ + struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur); + + if (*batchp == NULL) + return; + pte_free_submit(*batchp); + *batchp = NULL; +} -- cgit v1.2.3