[SPARC]: First cut of hugetlb on sparc64, 4MB currently.

arch/sparc64/Config.help

@@ -25,6 +25,14 @@ CONFIG_SMP
 
   If you don't know what to do here, say N.
 
+CONFIG_HUGETLB_PAGE
+  This enables support for huge pages.  User space applications
+  can make use of this support with the sys_alloc_hugepages and
+  sys_free_hugepages system calls.  If your applications are
+  huge page aware, then say Y here.
+
+  Otherwise, say N.
+
 CONFIG_PREEMPT
   This option reduces the latency of the kernel when reacting to
   real-time or interactive events by allowing a low priority process to

arch/sparc64/config.in

@@ -15,6 +15,8 @@ define_bool CONFIG_VT y
 define_bool CONFIG_VT_CONSOLE y
 define_bool CONFIG_HW_CONSOLE y
 
+bool 'SPARC64 Huge TLB Page Support' CONFIG_HUGETLB_PAGE
+
 bool 'Symmetric multi-processing support' CONFIG_SMP
 bool 'Preemptible Kernel' CONFIG_PREEMPT
 

arch/sparc64/kernel/sys_sparc.c

@@ -681,3 +681,100 @@ sys_perfctr(int opcode, unsigned long arg0, unsigned long arg1, unsigned long ar
 	};
 	return err;
 }
+
+#ifdef CONFIG_HUGETLB_PAGE
+#define HPAGE_ALIGN(x)  (((unsigned long)x + (HPAGE_SIZE -1)) & HPAGE_MASK)
+extern long     sys_munmap(unsigned long, size_t);
+
+/* get_addr function gets the currently unused virtual range in
+ * the current process's address space.  It returns the LARGE_PAGE_SIZE
+ * aligned address (in cases of success).  Other kernel generic
+ * routines only could gurantee that allocated address is PAGE_SIZE aligned.
+ */
+static long get_addr(unsigned long addr, unsigned long len)
+{
+	struct vm_area_struct   *vma;
+	if (addr) {
+		addr = HPAGE_ALIGN(addr);
+		vma = find_vma(current->mm, addr);
+		if (((TASK_SIZE - len) >= addr) &&
+				(!vma || addr + len <= vma->vm_start))
+			goto found_addr;
+	}
+	addr = HPAGE_ALIGN(TASK_UNMAPPED_BASE);
+	for (vma = find_vma(current->mm, addr); ; vma = vma->vm_next) {
+		if (TASK_SIZE - len < addr)
+			return -ENOMEM;
+		if (!vma || ((addr + len) < vma->vm_start))
+			goto found_addr;
+		addr = vma->vm_end;
+	}
+found_addr:
+	addr = HPAGE_ALIGN(addr);
+	return addr;
+}
+
+extern int alloc_hugetlb_pages(int, unsigned long, unsigned long, int, int);
+
+asmlinkage long
+sys_alloc_hugepages(int key, unsigned long addr, unsigned long len, int prot, int flag)
+{
+	struct mm_struct *mm = current->mm;
+	unsigned long raddr;
+	int retval;
+
+	if (key < 0)
+		return -EINVAL;
+	if (len & (HPAGE_SIZE - 1))
+		return -EINVAL;
+	down_write(&mm->mmap_sem);
+	raddr = get_addr(addr, len);
+	retval = 0;
+	if (raddr == -ENOMEM) {
+		retval = -ENOMEM;
+		goto raddr_out;
+	}
+	retval = alloc_hugetlb_pages(key, raddr, len, prot, flag);
+
+raddr_out:
+	up_write(&mm->mmap_sem);
+	if (retval < 0)
+		return (long) retval;
+
+	return raddr;
+}
+
+extern int free_hugepages(struct vm_area_struct *);
+
+asmlinkage int
+sys_free_hugepages(unsigned long addr)
+{
+	struct mm_struct *mm = current->mm;
+	struct vm_area_struct *vma;
+	int retval;
+
+	vma = find_vma(current->mm, addr);
+	if ((!vma) || (!is_vm_hugetlb_page(vma)) || (vma->vm_start!=addr))
+		return -EINVAL;
+	down_write(&mm->mmap_sem);
+	spin_lock(&mm->page_table_lock);
+	retval = free_hugepages(vma);
+	spin_unlock(&mm->page_table_lock);
+	up_write(&mm->mmap_sem);
+	return retval;
+}
+
+#else
+
+asmlinkage long
+sys_alloc_hugepages(int key, unsigned long addr, size_t len, int prot, int flag)
+{
+	return -ENOSYS;
+}
+asmlinkage int
+sys_free_hugepages(unsigned long addr)
+{
+	return -ENOSYS;
+}
+
+#endif

arch/sparc64/kernel/systbls.S

@@ -65,8 +65,8 @@ sys_call_table32:
 	.word sys32_ipc, sys32_sigreturn, sys_clone, sys_nis_syscall, sys32_adjtimex
 /*220*/	.word sys32_sigprocmask, sys32_create_module, sys32_delete_module, sys32_get_kernel_syms, sys_getpgid
 	.word sys32_bdflush, sys32_sysfs, sys_nis_syscall, sys32_setfsuid16, sys32_setfsgid16
-/*230*/	.word sys32_select, sys_time, sys_nis_syscall, sys_stime, sys_nis_syscall
-	.word sys_nis_syscall, sys_llseek, sys_mlock, sys_munlock, sys_mlockall
+/*230*/	.word sys32_select, sys_time, sys_nis_syscall, sys_stime, sys_alloc_hugepages
+	.word sys_free_hugepages, sys_llseek, sys_mlock, sys_munlock, sys_mlockall
 /*240*/	.word sys_munlockall, sys_sched_setparam, sys_sched_getparam, sys_sched_setscheduler, sys_sched_getscheduler
 	.word sys_sched_yield, sys_sched_get_priority_max, sys_sched_get_priority_min, sys32_sched_rr_get_interval, sys32_nanosleep
 /*250*/	.word sys32_mremap, sys32_sysctl, sys_getsid, sys_fdatasync, sys32_nfsservctl

arch/sparc64/mm/Makefile

@@ -7,4 +7,6 @@ EXTRA_AFLAGS := -ansi
 O_TARGET := mm.o
 obj-y    := ultra.o fault.o init.o generic.o extable.o modutil.o
 
+obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
+
 include $(TOPDIR)/Rules.make

arch/sparc64/mm/init.c

@@ -1690,6 +1690,13 @@ static void __init taint_real_pages(void)
 	}
 }
 
+#ifdef CONFIG_HUGETLB_PAGE
+long htlbpagemem = 0;
+long htlbpage_max;
+long htlbzone_pages;
+extern struct list_head htlbpage_freelist;
+#endif
+
 void __init mem_init(void)
 {
 	unsigned long codepages, datapages, initpages;
@@ -1766,6 +1773,32 @@ void __init mem_init(void)
 
 	if (tlb_type == cheetah || tlb_type == cheetah_plus)
 		cheetah_ecache_flush_init();
+#ifdef CONFIG_HUGETLB_PAGE
+	{
+		long i, j;
+		struct page *page, *map;
+
+		/* For now reserve quarter for hugetlb_pages. */
+		htlbzone_pages = (num_physpages >> ((HPAGE_SHIFT - PAGE_SHIFT) + 2)) ;
+
+		/* Will make this kernel command line. */
+		INIT_LIST_HEAD(&htlbpage_freelist);
+		for (i = 0; i < htlbzone_pages; i++) {
+			page = alloc_pages(GFP_ATOMIC, HUGETLB_PAGE_ORDER);
+			if (page == NULL)
+				break;
+			map = page;
+			for (j = 0; j < (HPAGE_SIZE / PAGE_SIZE); j++) {
+				SetPageReserved(map);
+				map++;
+			}
+			list_add(&page->list, &htlbpage_freelist);
+		}
+		printk("Total Huge_TLB_Page memory pages allocated %ld\n", i);
+		htlbzone_pages = htlbpagemem = i;
+		htlbpage_max = i;
+	}
+#endif
 }
 
 void free_initmem (void)

include/asm-sparc/unistd.h

@@ -249,8 +249,8 @@
 #define __NR_time               231 /* Linux Specific                              */
 /* #define __NR_oldstat         232    Linux Specific                              */
 #define __NR_stime              233 /* Linux Specific                              */
-/* #define __NR_oldfstat        234    Linux Specific                              */
-/* #define __NR_phys            235    Linux Specific                              */
+#define __NR_alloc_hugepages    234 /* Linux Specific                              */
+#define __NR_free_hugepages     235 /* Linux Specific                              */
 #define __NR__llseek            236 /* Linux Specific                              */
 #define __NR_mlock              237
 #define __NR_munlock            238

include/asm-sparc64/page.h

@@ -3,6 +3,8 @@
 #ifndef _SPARC64_PAGE_H
 #define _SPARC64_PAGE_H
 
+#include <linux/config.h>
+
 #define PAGE_SHIFT   13
 #ifndef __ASSEMBLY__
 /* I have my suspicions... -DaveM */
@@ -99,6 +101,14 @@ typedef unsigned long iopgprot_t;
 
 #endif /* (STRICT_MM_TYPECHECKS) */
 
+#define HPAGE_SHIFT		22
+
+#ifdef CONFIG_HUGETLB_PAGE
+#define HPAGE_SIZE		((1UL) << HPAGE_SHIFT)
+#define HPAGE_MASK		(~(HPAGE_SIZE - 1UL))
+#define HUGETLB_PAGE_ORDER	(HPAGE_SHIFT - PAGE_SHIFT)
+#endif
+
 #define TASK_UNMAPPED_BASE	(test_thread_flag(TIF_32BIT) ? \
 				 (0x0000000070000000UL) : (PAGE_OFFSET))
 

include/asm-sparc64/unistd.h

@@ -251,8 +251,8 @@
 #endif
 /* #define __NR_oldstat         232    Linux Specific                              */
 #define __NR_stime              233 /* Linux Specific                              */
-/* #define __NR_oldfstat        234    Linux Specific                              */
-/* #define __NR_phys            235    Linux Specific                              */
+#define __NR_alloc_hugepages    234 /* Linux Specific                              */
+#define __NR_free_hugepages     235 /* Linux Specific                              */
 #define __NR__llseek            236 /* Linux Specific                              */
 #define __NR_mlock              237
 #define __NR_munlock            238