linux_dsm_epyc7002/arch/x86/kernel/sys_x86_64.c
Borislav Petkov dfb09f9b7a x86, amd: Avoid cache aliasing penalties on AMD family 15h
This patch provides performance tuning for the "Bulldozer" CPU. With its
shared instruction cache there is a chance of generating an excessive
number of cache cross-invalidates when running specific workloads on the
cores of a compute module.

This excessive amount of cross-invalidations can be observed if cache
lines backed by shared physical memory alias in bits [14:12] of their
virtual addresses, as those bits are used for the index generation.

This patch addresses the issue by clearing all the bits in the [14:12]
slice of the file mapping's virtual address at generation time, thus
forcing those bits the same for all mappings of a single shared library
across processes and, in doing so, avoids instruction cache aliases.

It also adds the command line option "align_va_addr=(32|64|on|off)" with
which virtual address alignment can be enabled for 32-bit or 64-bit x86
individually, or both, or be completely disabled.

This change leaves virtual region address allocation on other families
and/or vendors unaffected.

Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1312550110-24160-2-git-send-email-bp@amd64.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
2011-08-05 12:26:44 -07:00

287 lines
6.8 KiB
C

#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 <asm/ia32.h>
#include <asm/syscalls.h>
struct __read_mostly va_alignment va_align = {
.flags = -1,
};
/*
* Align a virtual address to avoid aliasing in the I$ on AMD F15h.
*
* @flags denotes the allocation direction - bottomup or topdown -
* or vDSO; see call sites below.
*/
unsigned long align_addr(unsigned long addr, struct file *filp,
enum align_flags flags)
{
unsigned long tmp_addr;
/* handle 32- and 64-bit case with a single conditional */
if (va_align.flags < 0 || !(va_align.flags & (2 - mmap_is_ia32())))
return addr;
if (!(current->flags & PF_RANDOMIZE))
return addr;
if (!((flags & ALIGN_VDSO) || filp))
return addr;
tmp_addr = addr;
/*
* We need an address which is <= than the original
* one only when in topdown direction.
*/
if (!(flags & ALIGN_TOPDOWN))
tmp_addr += va_align.mask;
tmp_addr &= ~va_align.mask;
return tmp_addr;
}
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_IA32) && (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 = TASK_UNMAPPED_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;
unsigned long start_addr;
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;
}
if (((flags & MAP_32BIT) || test_thread_flag(TIF_IA32))
&& len <= mm->cached_hole_size) {
mm->cached_hole_size = 0;
mm->free_area_cache = begin;
}
addr = mm->free_area_cache;
if (addr < begin)
addr = begin;
start_addr = addr;
full_search:
addr = align_addr(addr, filp, 0);
for (vma = find_vma(mm, addr); ; vma = vma->vm_next) {
/* At this point: (!vma || addr < vma->vm_end). */
if (end - len < addr) {
/*
* Start a new search - just in case we missed
* some holes.
*/
if (start_addr != begin) {
start_addr = addr = begin;
mm->cached_hole_size = 0;
goto full_search;
}
return -ENOMEM;
}
if (!vma || addr + len <= vma->vm_start) {
/*
* Remember the place where we stopped the search:
*/
mm->free_area_cache = addr + len;
return addr;
}
if (addr + mm->cached_hole_size < vma->vm_start)
mm->cached_hole_size = vma->vm_start - addr;
addr = vma->vm_end;
addr = align_addr(addr, filp, 0);
}
}
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;
/* 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 legact mmap base */
if (!test_thread_flag(TIF_IA32) && (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;
}
/* check if free_area_cache is useful for us */
if (len <= mm->cached_hole_size) {
mm->cached_hole_size = 0;
mm->free_area_cache = mm->mmap_base;
}
/* either no address requested or can't fit in requested address hole */
addr = mm->free_area_cache;
/* make sure it can fit in the remaining address space */
if (addr > len) {
unsigned long tmp_addr = align_addr(addr - len, filp,
ALIGN_TOPDOWN);
vma = find_vma(mm, tmp_addr);
if (!vma || tmp_addr + len <= vma->vm_start)
/* remember the address as a hint for next time */
return mm->free_area_cache = tmp_addr;
}
if (mm->mmap_base < len)
goto bottomup;
addr = mm->mmap_base-len;
do {
addr = align_addr(addr, filp, ALIGN_TOPDOWN);
/*
* Lookup failure means no vma is above this address,
* else if new region fits below vma->vm_start,
* return with success:
*/
vma = find_vma(mm, addr);
if (!vma || addr+len <= vma->vm_start)
/* remember the address as a hint for next time */
return mm->free_area_cache = addr;
/* remember the largest hole we saw so far */
if (addr + mm->cached_hole_size < vma->vm_start)
mm->cached_hole_size = vma->vm_start - addr;
/* try just below the current vma->vm_start */
addr = vma->vm_start-len;
} while (len < vma->vm_start);
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.
*/
mm->cached_hole_size = ~0UL;
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 = mm->mmap_base;
mm->cached_hole_size = ~0UL;
return addr;
}