1
linux/arch/arm/mm/fault-armv.c
Nicolas Pitre 08e445bd6a [ARM] 5366/1: fix shared memory coherency with VIVT L1 + L2 caches
When there are multiple L1-aliasing userland mappings of the same physical
page, we currently remap each of them uncached, to prevent VIVT cache
aliasing issues. (E.g. writes to one of the mappings not being immediately
visible via another mapping.)  However, when we do this remapping, there
could still be stale data in the L2 cache, and an uncached mapping might
bypass L2 and go straight to RAM.  This would cause reads from such
mappings to see old data (until the dirty L2 line is eventually evicted.)

This issue is solved by forcing a L2 cache flush whenever the shared page
is made L1 uncacheable.

Ideally, we would make L1 uncacheable and L2 cacheable as L2 is PIPT. But
Feroceon does not support that combination, and the TEX=5 C=0 B=0 encoding
for XSc3 doesn't appear to work in practice.

Signed-off-by: Nicolas Pitre <nico@marvell.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2009-01-28 16:55:00 +00:00

232 lines
5.6 KiB
C

/*
* linux/arch/arm/mm/fault-armv.c
*
* Copyright (C) 1995 Linus Torvalds
* Modifications for ARM processor (c) 1995-2002 Russell King
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*/
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/bitops.h>
#include <linux/vmalloc.h>
#include <linux/init.h>
#include <linux/pagemap.h>
#include <asm/bugs.h>
#include <asm/cacheflush.h>
#include <asm/cachetype.h>
#include <asm/pgtable.h>
#include <asm/tlbflush.h>
static unsigned long shared_pte_mask = L_PTE_MT_BUFFERABLE;
/*
* We take the easy way out of this problem - we make the
* PTE uncacheable. However, we leave the write buffer on.
*
* Note that the pte lock held when calling update_mmu_cache must also
* guard the pte (somewhere else in the same mm) that we modify here.
* Therefore those configurations which might call adjust_pte (those
* without CONFIG_CPU_CACHE_VIPT) cannot support split page_table_lock.
*/
static int adjust_pte(struct vm_area_struct *vma, unsigned long address)
{
pgd_t *pgd;
pmd_t *pmd;
pte_t *pte, entry;
int ret;
pgd = pgd_offset(vma->vm_mm, address);
if (pgd_none(*pgd))
goto no_pgd;
if (pgd_bad(*pgd))
goto bad_pgd;
pmd = pmd_offset(pgd, address);
if (pmd_none(*pmd))
goto no_pmd;
if (pmd_bad(*pmd))
goto bad_pmd;
pte = pte_offset_map(pmd, address);
entry = *pte;
/*
* If this page is present, it's actually being shared.
*/
ret = pte_present(entry);
/*
* If this page isn't present, or is already setup to
* fault (ie, is old), we can safely ignore any issues.
*/
if (ret && (pte_val(entry) & L_PTE_MT_MASK) != shared_pte_mask) {
unsigned long pfn = pte_pfn(entry);
flush_cache_page(vma, address, pfn);
outer_flush_range((pfn << PAGE_SHIFT),
(pfn << PAGE_SHIFT) + PAGE_SIZE);
pte_val(entry) &= ~L_PTE_MT_MASK;
pte_val(entry) |= shared_pte_mask;
set_pte_at(vma->vm_mm, address, pte, entry);
flush_tlb_page(vma, address);
}
pte_unmap(pte);
return ret;
bad_pgd:
pgd_ERROR(*pgd);
pgd_clear(pgd);
no_pgd:
return 0;
bad_pmd:
pmd_ERROR(*pmd);
pmd_clear(pmd);
no_pmd:
return 0;
}
static void
make_coherent(struct address_space *mapping, struct vm_area_struct *vma, unsigned long addr, unsigned long pfn)
{
struct mm_struct *mm = vma->vm_mm;
struct vm_area_struct *mpnt;
struct prio_tree_iter iter;
unsigned long offset;
pgoff_t pgoff;
int aliases = 0;
pgoff = vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT);
/*
* If we have any shared mappings that are in the same mm
* space, then we need to handle them specially to maintain
* cache coherency.
*/
flush_dcache_mmap_lock(mapping);
vma_prio_tree_foreach(mpnt, &iter, &mapping->i_mmap, pgoff, pgoff) {
/*
* If this VMA is not in our MM, we can ignore it.
* Note that we intentionally mask out the VMA
* that we are fixing up.
*/
if (mpnt->vm_mm != mm || mpnt == vma)
continue;
if (!(mpnt->vm_flags & VM_MAYSHARE))
continue;
offset = (pgoff - mpnt->vm_pgoff) << PAGE_SHIFT;
aliases += adjust_pte(mpnt, mpnt->vm_start + offset);
}
flush_dcache_mmap_unlock(mapping);
if (aliases)
adjust_pte(vma, addr);
else
flush_cache_page(vma, addr, pfn);
}
/*
* Take care of architecture specific things when placing a new PTE into
* a page table, or changing an existing PTE. Basically, there are two
* things that we need to take care of:
*
* 1. If PG_dcache_dirty is set for the page, we need to ensure
* that any cache entries for the kernels virtual memory
* range are written back to the page.
* 2. If we have multiple shared mappings of the same space in
* an object, we need to deal with the cache aliasing issues.
*
* Note that the pte lock will be held.
*/
void update_mmu_cache(struct vm_area_struct *vma, unsigned long addr, pte_t pte)
{
unsigned long pfn = pte_pfn(pte);
struct address_space *mapping;
struct page *page;
if (!pfn_valid(pfn))
return;
page = pfn_to_page(pfn);
mapping = page_mapping(page);
if (mapping) {
#ifndef CONFIG_SMP
int dirty = test_and_clear_bit(PG_dcache_dirty, &page->flags);
if (dirty)
__flush_dcache_page(mapping, page);
#endif
if (cache_is_vivt())
make_coherent(mapping, vma, addr, pfn);
else if (vma->vm_flags & VM_EXEC)
__flush_icache_all();
}
}
/*
* Check whether the write buffer has physical address aliasing
* issues. If it has, we need to avoid them for the case where
* we have several shared mappings of the same object in user
* space.
*/
static int __init check_writebuffer(unsigned long *p1, unsigned long *p2)
{
register unsigned long zero = 0, one = 1, val;
local_irq_disable();
mb();
*p1 = one;
mb();
*p2 = zero;
mb();
val = *p1;
mb();
local_irq_enable();
return val != zero;
}
void __init check_writebuffer_bugs(void)
{
struct page *page;
const char *reason;
unsigned long v = 1;
printk(KERN_INFO "CPU: Testing write buffer coherency: ");
page = alloc_page(GFP_KERNEL);
if (page) {
unsigned long *p1, *p2;
pgprot_t prot = __pgprot(L_PTE_PRESENT|L_PTE_YOUNG|
L_PTE_DIRTY|L_PTE_WRITE|
L_PTE_MT_BUFFERABLE);
p1 = vmap(&page, 1, VM_IOREMAP, prot);
p2 = vmap(&page, 1, VM_IOREMAP, prot);
if (p1 && p2) {
v = check_writebuffer(p1, p2);
reason = "enabling work-around";
} else {
reason = "unable to map memory\n";
}
vunmap(p1);
vunmap(p2);
put_page(page);
} else {
reason = "unable to grab page\n";
}
if (v) {
printk("failed, %s\n", reason);
shared_pte_mask = L_PTE_MT_UNCACHED;
} else {
printk("ok\n");
}
}