// SPDX-License-Identifier: GPL-2.0-only
/*
* linux / arch / arm / mm / fault - armv . c
*
* Copyright ( C ) 1995 Linus Torvalds
* Modifications for ARM processor ( c ) 1995 - 2002 Russell King
*/
#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 <linux/gfp.h>
#include <asm /bugs.h>
#include <asm /cacheflush.h>
#include <asm /cachetype.h>
#include <asm /tlbflush.h>
#include "mm.h"
static pteval_t shared_pte_mask = L_PTE_MT_BUFFERABLE;
#if __LINUX_ARM_ARCH__ < 6
/*
* 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 do_adjust_pte(struct vm_area_struct *vma, unsigned long address,
unsigned long pfn, pte_t *ptep)
{
pte_t entry = *ptep;
int ret;
/*
* 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) {
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, ptep, entry);
flush_tlb_page(vma, address);
}
return ret;
}
static int adjust_pte(struct vm_area_struct *vma, unsigned long address,
unsigned long pfn, bool need_lock)
{
spinlock_t *ptl;
pgd_t *pgd;
p4d_t *p4d;
pud_t *pud;
pmd_t *pmd;
pte_t *pte;
pmd_t pmdval;
int ret;
pgd = pgd_offset(vma->vm_mm, address);
if (pgd_none_or_clear_bad(pgd))
return 0 ;
p4d = p4d_offset(pgd, address);
if (p4d_none_or_clear_bad(p4d))
return 0 ;
pud = pud_offset(p4d, address);
if (pud_none_or_clear_bad(pud))
return 0 ;
pmd = pmd_offset(pud, address);
if (pmd_none_or_clear_bad(pmd))
return 0 ;
again:
/*
* This is called while another page table is mapped , so we
* must use the nested version . This also means we need to
* open - code the spin - locking .
*/
pte = pte_offset_map_rw_nolock(vma->vm_mm, pmd, address, &pmdval, &ptl);
if (!pte)
return 0 ;
if (need_lock) {
/*
* Use nested version here to indicate that we are already
* holding one similar spinlock .
*/
spin_lock_nested(ptl, SINGLE_DEPTH_NESTING);
if (unlikely(!pmd_same(pmdval, pmdp_get_lockless(pmd)))) {
pte_unmap_unlock(pte, ptl);
goto again;
}
}
ret = do_adjust_pte(vma, address, pfn, pte);
if (need_lock)
spin_unlock(ptl);
pte_unmap(pte);
return ret;
}
static void
make_coherent(struct address_space *mapping, struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep, unsigned long pfn)
{
const unsigned long pmd_start_addr = ALIGN_DOWN(addr, PMD_SIZE);
const unsigned long pmd_end_addr = pmd_start_addr + PMD_SIZE;
struct mm_struct *mm = vma->vm_mm;
struct vm_area_struct *mpnt;
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_interval_tree_foreach(mpnt, &mapping->i_mmap, pgoff, pgoff) {
/*
* If we are using split PTE locks , then we need to take the pte
* lock . Otherwise we are using shared mm - > page_table_lock which
* is already locked , thus cannot take it .
*/
bool need_lock = IS_ENABLED(CONFIG_SPLIT_PTE_PTLOCKS);
unsigned long mpnt_addr;
/*
* 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;
mpnt_addr = mpnt->vm_start + offset;
/* Avoid deadlocks by not grabbing the same PTE lock again. */
if (mpnt_addr >= pmd_start_addr && mpnt_addr < pmd_end_addr)
need_lock = false ;
aliases += adjust_pte(mpnt, mpnt_addr, pfn, need_lock);
}
flush_dcache_mmap_unlock(mapping);
if (aliases)
do_adjust_pte(vma, addr, pfn, ptep);
}
/*
* 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_clean is not 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_range(struct vm_fault *vmf, struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep, unsigned int nr)
{
unsigned long pfn = pte_pfn(*ptep);
struct address_space *mapping;
struct folio *folio;
if (!pfn_valid(pfn))
return ;
/*
* The zero page is never written to , so never has any dirty
* cache lines , and therefore never needs to be flushed .
*/
if (is_zero_pfn(pfn))
return ;
folio = page_folio(pfn_to_page(pfn));
mapping = folio_flush_mapping(folio);
if (!test_and_set_bit(PG_dcache_clean, &folio->flags))
__flush_dcache_folio(mapping, folio);
if (mapping) {
if (cache_is_vivt())
make_coherent(mapping, vma, addr, ptep, pfn);
else if (vma->vm_flags & VM_EXEC)
__flush_icache_all();
}
}
#endif /* __LINUX_ARM_ARCH__ < 6 */
/*
* 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 ;
pr_info("CPU: Testing write buffer coherency: " );
page = alloc_page(GFP_KERNEL);
if (page) {
unsigned long *p1, *p2;
pgprot_t prot = __pgprot_modify(PAGE_KERNEL,
L_PTE_MT_MASK, 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) {
pr_cont("failed, %s\n" , reason);
shared_pte_mask = L_PTE_MT_UNCACHED;
} else {
pr_cont("ok\n" );
}
}
Messung V0.5 in Prozent C=95 H=92 G=93
¤ Dauer der Verarbeitung: 0.12 Sekunden
(vorverarbeitet am 2026-09-28)
¤
*© Formatika GbR, Deutschland