Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / arch / x86 / kernel / sys_x86_64.c
diff --git a/kernel/arch/x86/kernel/sys_x86_64.c b/kernel/arch/x86/kernel/sys_x86_64.c
new file mode 100644 (file)
index 0000000..10e0272
--- /dev/null
@@ -0,0 +1,216 @@
+#include <linux/errno.h>
+#include <linux/sched.h>
+#include <linux/syscalls.h>
+#include <linux/mm.h>
+#include <linux/fs.h>
+#include <linux/smp.h>
+#include <linux/sem.h>
+#include <linux/msg.h>
+#include <linux/shm.h>
+#include <linux/stat.h>
+#include <linux/mman.h>
+#include <linux/file.h>
+#include <linux/utsname.h>
+#include <linux/personality.h>
+#include <linux/random.h>
+#include <linux/uaccess.h>
+#include <linux/elf.h>
+
+#include <asm/ia32.h>
+#include <asm/syscalls.h>
+
+/*
+ * Align a virtual address to avoid aliasing in the I$ on AMD F15h.
+ */
+static unsigned long get_align_mask(void)
+{
+       /* handle 32- and 64-bit case with a single conditional */
+       if (va_align.flags < 0 || !(va_align.flags & (2 - mmap_is_ia32())))
+               return 0;
+
+       if (!(current->flags & PF_RANDOMIZE))
+               return 0;
+
+       return va_align.mask;
+}
+
+/*
+ * To avoid aliasing in the I$ on AMD F15h, the bits defined by the
+ * va_align.bits, [12:upper_bit), are set to a random value instead of
+ * zeroing them. This random value is computed once per boot. This form
+ * of ASLR is known as "per-boot ASLR".
+ *
+ * To achieve this, the random value is added to the info.align_offset
+ * value before calling vm_unmapped_area() or ORed directly to the
+ * address.
+ */
+static unsigned long get_align_bits(void)
+{
+       return va_align.bits & get_align_mask();
+}
+
+unsigned long align_vdso_addr(unsigned long addr)
+{
+       unsigned long align_mask = get_align_mask();
+       addr = (addr + align_mask) & ~align_mask;
+       return addr | get_align_bits();
+}
+
+static int __init control_va_addr_alignment(char *str)
+{
+       /* guard against enabling this on other CPU families */
+       if (va_align.flags < 0)
+               return 1;
+
+       if (*str == 0)
+               return 1;
+
+       if (*str == '=')
+               str++;
+
+       if (!strcmp(str, "32"))
+               va_align.flags = ALIGN_VA_32;
+       else if (!strcmp(str, "64"))
+               va_align.flags = ALIGN_VA_64;
+       else if (!strcmp(str, "off"))
+               va_align.flags = 0;
+       else if (!strcmp(str, "on"))
+               va_align.flags = ALIGN_VA_32 | ALIGN_VA_64;
+       else
+               return 0;
+
+       return 1;
+}
+__setup("align_va_addr", control_va_addr_alignment);
+
+SYSCALL_DEFINE6(mmap, unsigned long, addr, unsigned long, len,
+               unsigned long, prot, unsigned long, flags,
+               unsigned long, fd, unsigned long, off)
+{
+       long error;
+       error = -EINVAL;
+       if (off & ~PAGE_MASK)
+               goto out;
+
+       error = sys_mmap_pgoff(addr, len, prot, flags, fd, off >> PAGE_SHIFT);
+out:
+       return error;
+}
+
+static void find_start_end(unsigned long flags, unsigned long *begin,
+                          unsigned long *end)
+{
+       if (!test_thread_flag(TIF_ADDR32) && (flags & MAP_32BIT)) {
+               unsigned long new_begin;
+               /* This is usually used needed to map code in small
+                  model, so it needs to be in the first 31bit. Limit
+                  it to that.  This means we need to move the
+                  unmapped base down for this case. This can give
+                  conflicts with the heap, but we assume that glibc
+                  malloc knows how to fall back to mmap. Give it 1GB
+                  of playground for now. -AK */
+               *begin = 0x40000000;
+               *end = 0x80000000;
+               if (current->flags & PF_RANDOMIZE) {
+                       new_begin = randomize_range(*begin, *begin + 0x02000000, 0);
+                       if (new_begin)
+                               *begin = new_begin;
+               }
+       } else {
+               *begin = current->mm->mmap_legacy_base;
+               *end = TASK_SIZE;
+       }
+}
+
+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;
+       struct vm_unmapped_area_info info;
+       unsigned long begin, end;
+
+       if (flags & MAP_FIXED)
+               return addr;
+
+       find_start_end(flags, &begin, &end);
+
+       if (len > end)
+               return -ENOMEM;
+
+       if (addr) {
+               addr = PAGE_ALIGN(addr);
+               vma = find_vma(mm, addr);
+               if (end - len >= addr &&
+                   (!vma || addr + len <= vma->vm_start))
+                       return addr;
+       }
+
+       info.flags = 0;
+       info.length = len;
+       info.low_limit = begin;
+       info.high_limit = end;
+       info.align_mask = 0;
+       info.align_offset = pgoff << PAGE_SHIFT;
+       if (filp) {
+               info.align_mask = get_align_mask();
+               info.align_offset += get_align_bits();
+       }
+       return vm_unmapped_area(&info);
+}
+
+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;
+       struct mm_struct *mm = current->mm;
+       unsigned long addr = addr0;
+       struct vm_unmapped_area_info info;
+
+       /* requested length too big for entire address space */
+       if (len > TASK_SIZE)
+               return -ENOMEM;
+
+       if (flags & MAP_FIXED)
+               return addr;
+
+       /* for MAP_32BIT mappings we force the legacy mmap base */
+       if (!test_thread_flag(TIF_ADDR32) && (flags & MAP_32BIT))
+               goto bottomup;
+
+       /* 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))
+                       return addr;
+       }
+
+       info.flags = VM_UNMAPPED_AREA_TOPDOWN;
+       info.length = len;
+       info.low_limit = PAGE_SIZE;
+       info.high_limit = mm->mmap_base;
+       info.align_mask = 0;
+       info.align_offset = pgoff << PAGE_SHIFT;
+       if (filp) {
+               info.align_mask = get_align_mask();
+               info.align_offset += get_align_bits();
+       }
+       addr = vm_unmapped_area(&info);
+       if (!(addr & ~PAGE_MASK))
+               return addr;
+       VM_BUG_ON(addr != -ENOMEM);
+
+bottomup:
+       /*
+        * 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.
+        */
+       return arch_get_unmapped_area(filp, addr0, len, pgoff, flags);
+}