#include <asm/uaccess.h>
#include <asm/pgalloc.h>
+/*
+ * The current flushing context - we pass it instead of 5 arguments:
+ */
+struct cpa_data {
+ unsigned long vaddr;
+ pgprot_t mask_set;
+ pgprot_t mask_clr;
+ int numpages;
+ int flushtlb;
+};
+
static inline int
within(unsigned long addr, unsigned long start, unsigned long end)
{
/*
* Flushing functions
*/
-void clflush_cache_range(void *addr, int size)
+
+/**
+ * clflush_cache_range - flush a cache range with clflush
+ * @addr: virtual start address
+ * @size: number of bytes to flush
+ *
+ * clflush is an unordered instruction which needs fencing with mfence
+ * to avoid ordering issues.
+ */
+void clflush_cache_range(void *vaddr, unsigned int size)
{
- int i;
+ void *vend = vaddr + size - 1;
+
+ mb();
+
+ for (; vaddr < vend; vaddr += boot_cpu_data.x86_clflush_size)
+ clflush(vaddr);
+ /*
+ * Flush any possible final partial cacheline:
+ */
+ clflush(vend);
- for (i = 0; i < size; i += boot_cpu_data.x86_clflush_size)
- clflush(addr+i);
+ mb();
}
-static void flush_kernel_map(void *arg)
+static void __cpa_flush_all(void *arg)
{
+ unsigned long cache = (unsigned long)arg;
+
/*
* Flush all to work around Errata in early athlons regarding
* large page flushing.
*/
__flush_tlb_all();
- if (boot_cpu_data.x86_model >= 4)
+ if (cache && boot_cpu_data.x86_model >= 4)
wbinvd();
}
-static void global_flush_tlb(void)
+static void cpa_flush_all(unsigned long cache)
+{
+ BUG_ON(irqs_disabled());
+
+ on_each_cpu(__cpa_flush_all, (void *) cache, 1, 1);
+}
+
+static void __cpa_flush_range(void *arg)
+{
+ /*
+ * We could optimize that further and do individual per page
+ * tlb invalidates for a low number of pages. Caveat: we must
+ * flush the high aliases on 64bit as well.
+ */
+ __flush_tlb_all();
+}
+
+static void cpa_flush_range(unsigned long start, int numpages, int cache)
{
+ unsigned int i, level;
+ unsigned long addr;
+
BUG_ON(irqs_disabled());
+ WARN_ON(PAGE_ALIGN(start) != start);
+
+ on_each_cpu(__cpa_flush_range, NULL, 1, 1);
+
+ if (!cache)
+ return;
+
+ /*
+ * We only need to flush on one CPU,
+ * clflush is a MESI-coherent instruction that
+ * will cause all other CPUs to flush the same
+ * cachelines:
+ */
+ for (i = 0, addr = start; i < numpages; i++, addr += PAGE_SIZE) {
+ pte_t *pte = lookup_address(addr, &level);
+
+ /*
+ * Only flush present addresses:
+ */
+ if (pte && (pte_val(*pte) & _PAGE_PRESENT))
+ clflush_cache_range((void *) addr, PAGE_SIZE);
+ }
+}
- on_each_cpu(flush_kernel_map, NULL, 1, 1);
+#define HIGH_MAP_START __START_KERNEL_map
+#define HIGH_MAP_END (__START_KERNEL_map + KERNEL_TEXT_SIZE)
+
+
+/*
+ * Converts a virtual address to a X86-64 highmap address
+ */
+static unsigned long virt_to_highmap(void *address)
+{
+#ifdef CONFIG_X86_64
+ return __pa((unsigned long)address) + HIGH_MAP_START - phys_base;
+#else
+ return (unsigned long)address;
+#endif
}
/*
*/
if (within(address, (unsigned long)_text, (unsigned long)_etext))
pgprot_val(forbidden) |= _PAGE_NX;
+ /*
+ * Do the same for the x86-64 high kernel mapping
+ */
+ if (within(address, virt_to_highmap(_text), virt_to_highmap(_etext)))
+ pgprot_val(forbidden) |= _PAGE_NX;
-#ifdef CONFIG_DEBUG_RODATA
/* The .rodata section needs to be read-only */
if (within(address, (unsigned long)__start_rodata,
(unsigned long)__end_rodata))
pgprot_val(forbidden) |= _PAGE_RW;
-#endif
+ /*
+ * Do the same for the x86-64 high kernel mapping
+ */
+ if (within(address, virt_to_highmap(__start_rodata),
+ virt_to_highmap(__end_rodata)))
+ pgprot_val(forbidden) |= _PAGE_RW;
prot = __pgprot(pgprot_val(prot) & ~pgprot_val(forbidden));
return prot;
}
+/*
+ * Lookup the page table entry for a virtual address. Return a pointer
+ * to the entry and the level of the mapping.
+ *
+ * Note: We return pud and pmd either when the entry is marked large
+ * or when the present bit is not set. Otherwise we would return a
+ * pointer to a nonexisting mapping.
+ */
pte_t *lookup_address(unsigned long address, int *level)
{
pgd_t *pgd = pgd_offset_k(address);
if (pgd_none(*pgd))
return NULL;
+
pud = pud_offset(pgd, address);
if (pud_none(*pud))
return NULL;
+
+ *level = PG_LEVEL_1G;
+ if (pud_large(*pud) || !pud_present(*pud))
+ return (pte_t *)pud;
+
pmd = pmd_offset(pud, address);
if (pmd_none(*pmd))
return NULL;
*level = PG_LEVEL_2M;
- if (pmd_large(*pmd))
+ if (pmd_large(*pmd) || !pmd_present(*pmd))
return (pte_t *)pmd;
*level = PG_LEVEL_4K;
+
return pte_offset_kernel(pmd, address);
}
+/*
+ * Set the new pmd in all the pgds we know about:
+ */
static void __set_pmd_pte(pte_t *kpte, unsigned long address, pte_t pte)
{
/* change init_mm */
if (!SHARED_KERNEL_PMD) {
struct page *page;
- for (page = pgd_list; page; page = (struct page *)page->index) {
+ list_for_each_entry(page, &pgd_list, lru) {
pgd_t *pgd;
pud_t *pud;
pmd_t *pmd;
#endif
}
+static int
+try_preserve_large_page(pte_t *kpte, unsigned long address,
+ struct cpa_data *cpa)
+{
+ unsigned long nextpage_addr, numpages, pmask, psize, flags;
+ pte_t new_pte, old_pte, *tmp;
+ pgprot_t old_prot, new_prot;
+ int level, do_split = 1;
+
+ spin_lock_irqsave(&pgd_lock, flags);
+ /*
+ * Check for races, another CPU might have split this page
+ * up already:
+ */
+ tmp = lookup_address(address, &level);
+ if (tmp != kpte)
+ goto out_unlock;
+
+ switch (level) {
+ case PG_LEVEL_2M:
+ psize = PMD_PAGE_SIZE;
+ pmask = PMD_PAGE_MASK;
+ break;
+#ifdef CONFIG_X86_64
+ case PG_LEVEL_1G:
+ psize = PMD_PAGE_SIZE;
+ pmask = PMD_PAGE_MASK;
+ break;
+#endif
+ default:
+ do_split = -EINVAL;
+ goto out_unlock;
+ }
+
+ /*
+ * Calculate the number of pages, which fit into this large
+ * page starting at address:
+ */
+ nextpage_addr = (address + psize) & pmask;
+ numpages = (nextpage_addr - address) >> PAGE_SHIFT;
+ if (numpages < cpa->numpages)
+ cpa->numpages = numpages;
+
+ /*
+ * We are safe now. Check whether the new pgprot is the same:
+ */
+ old_pte = *kpte;
+ old_prot = new_prot = pte_pgprot(old_pte);
+
+ pgprot_val(new_prot) &= ~pgprot_val(cpa->mask_clr);
+ pgprot_val(new_prot) |= pgprot_val(cpa->mask_set);
+ new_prot = static_protections(new_prot, address);
+
+ /*
+ * If there are no changes, return. maxpages has been updated
+ * above:
+ */
+ if (pgprot_val(new_prot) == pgprot_val(old_prot)) {
+ do_split = 0;
+ goto out_unlock;
+ }
+
+ /*
+ * We need to change the attributes. Check, whether we can
+ * change the large page in one go. We request a split, when
+ * the address is not aligned and the number of pages is
+ * smaller than the number of pages in the large page. Note
+ * that we limited the number of possible pages already to
+ * the number of pages in the large page.
+ */
+ if (address == (nextpage_addr - psize) && cpa->numpages == numpages) {
+ /*
+ * The address is aligned and the number of pages
+ * covers the full page.
+ */
+ new_pte = pfn_pte(pte_pfn(old_pte), canon_pgprot(new_prot));
+ __set_pmd_pte(kpte, address, new_pte);
+ cpa->flushtlb = 1;
+ do_split = 0;
+ }
+
+out_unlock:
+ spin_unlock_irqrestore(&pgd_lock, flags);
+
+ return do_split;
+}
+
static int split_large_page(pte_t *kpte, unsigned long address)
{
- pgprot_t ref_prot = pte_pgprot(pte_clrhuge(*kpte));
+ unsigned long flags, pfn, pfninc = 1;
gfp_t gfp_flags = GFP_KERNEL;
- unsigned long flags;
- unsigned long addr;
+ unsigned int i, level;
pte_t *pbase, *tmp;
+ pgprot_t ref_prot;
struct page *base;
- int i, level;
#ifdef CONFIG_DEBUG_PAGEALLOC
- gfp_flags = GFP_ATOMIC;
+ gfp_flags = GFP_ATOMIC | __GFP_NOWARN;
#endif
base = alloc_pages(gfp_flags, 0);
if (!base)
* up for us already:
*/
tmp = lookup_address(address, &level);
- if (tmp != kpte) {
- WARN_ON_ONCE(1);
+ if (tmp != kpte)
goto out_unlock;
- }
- address = __pa(address);
- addr = address & LARGE_PAGE_MASK;
pbase = (pte_t *)page_address(base);
#ifdef CONFIG_X86_32
paravirt_alloc_pt(&init_mm, page_to_pfn(base));
#endif
+ ref_prot = pte_pgprot(pte_clrhuge(*kpte));
- for (i = 0; i < PTRS_PER_PTE; i++, addr += PAGE_SIZE)
- set_pte(&pbase[i], pfn_pte(addr >> PAGE_SHIFT, ref_prot));
+#ifdef CONFIG_X86_64
+ if (level == PG_LEVEL_1G) {
+ pfninc = PMD_PAGE_SIZE >> PAGE_SHIFT;
+ pgprot_val(ref_prot) |= _PAGE_PSE;
+ }
+#endif
+
+ /*
+ * Get the target pfn from the original entry:
+ */
+ pfn = pte_pfn(*kpte);
+ for (i = 0; i < PTRS_PER_PTE; i++, pfn += pfninc)
+ set_pte(&pbase[i], pfn_pte(pfn, ref_prot));
/*
- * Install the new, split up pagetable. Important detail here:
+ * Install the new, split up pagetable. Important details here:
*
* On Intel the NX bit of all levels must be cleared to make a
* page executable. See section 4.13.2 of Intel 64 and IA-32
* Architectures Software Developer's Manual).
+ *
+ * Mark the entry present. The current mapping might be
+ * set to not present, which we preserved above.
*/
ref_prot = pte_pgprot(pte_mkexec(pte_clrhuge(*kpte)));
+ pgprot_val(ref_prot) |= _PAGE_PRESENT;
__set_pmd_pte(kpte, address, mk_pte(base, ref_prot));
base = NULL;
return 0;
}
-static int
-__change_page_attr(unsigned long address, unsigned long pfn, pgprot_t prot)
+static int __change_page_attr(unsigned long address, struct cpa_data *cpa)
{
+ int level, do_split, err;
struct page *kpte_page;
- int level, err = 0;
pte_t *kpte;
-#ifdef CONFIG_X86_32
- BUG_ON(pfn > max_low_pfn);
-#endif
-
repeat:
kpte = lookup_address(address, &level);
if (!kpte)
BUG_ON(PageLRU(kpte_page));
BUG_ON(PageCompound(kpte_page));
- prot = static_protections(prot, address);
-
if (level == PG_LEVEL_4K) {
- WARN_ON_ONCE(pgprot_val(prot) & _PAGE_PSE);
- set_pte_atomic(kpte, pfn_pte(pfn, canon_pgprot(prot)));
- } else {
- /* Clear the PSE bit for the 4k level pages ! */
- pgprot_val(prot) = pgprot_val(prot) & ~_PAGE_PSE;
-
- err = split_large_page(kpte, address);
- if (!err)
- goto repeat;
+ pte_t new_pte, old_pte = *kpte;
+ pgprot_t new_prot = pte_pgprot(old_pte);
+
+ if(!pte_val(old_pte)) {
+ printk(KERN_WARNING "CPA: called for zero pte. "
+ "vaddr = %lx cpa->vaddr = %lx\n", address,
+ cpa->vaddr);
+ WARN_ON(1);
+ return -EINVAL;
+ }
+
+ pgprot_val(new_prot) &= ~pgprot_val(cpa->mask_clr);
+ pgprot_val(new_prot) |= pgprot_val(cpa->mask_set);
+
+ new_prot = static_protections(new_prot, address);
+
+ /*
+ * We need to keep the pfn from the existing PTE,
+ * after all we're only going to change it's attributes
+ * not the memory it points to
+ */
+ new_pte = pfn_pte(pte_pfn(old_pte), canon_pgprot(new_prot));
+
+ /*
+ * Do we really change anything ?
+ */
+ if (pte_val(old_pte) != pte_val(new_pte)) {
+ set_pte_atomic(kpte, new_pte);
+ cpa->flushtlb = 1;
+ }
+ cpa->numpages = 1;
+ return 0;
+ }
+
+ /*
+ * Check, whether we can keep the large page intact
+ * and just change the pte:
+ */
+ do_split = try_preserve_large_page(kpte, address, cpa);
+ /*
+ * When the range fits into the existing large page,
+ * return. cp->numpages and cpa->tlbflush have been updated in
+ * try_large_page:
+ */
+ if (do_split <= 0)
+ return do_split;
+
+ /*
+ * We have to split the large page:
+ */
+ err = split_large_page(kpte, address);
+ if (!err) {
+ cpa->flushtlb = 1;
+ goto repeat;
}
+
return err;
}
*
* Modules and drivers should use the set_memory_* APIs instead.
*/
-
-static int change_page_attr_addr(unsigned long address, pgprot_t prot)
+static int change_page_attr_addr(struct cpa_data *cpa)
{
- int err = 0, kernel_map = 0;
- unsigned long pfn = __pa(address) >> PAGE_SHIFT;
+ int err;
+ unsigned long address = cpa->vaddr;
#ifdef CONFIG_X86_64
- if (address >= __START_KERNEL_map &&
- address < __START_KERNEL_map + KERNEL_TEXT_SIZE) {
+ unsigned long phys_addr = __pa(address);
- address = (unsigned long)__va(__pa(address));
- kernel_map = 1;
- }
+ /*
+ * If we are inside the high mapped kernel range, then we
+ * fixup the low mapping first. __va() returns the virtual
+ * address in the linear mapping:
+ */
+ if (within(address, HIGH_MAP_START, HIGH_MAP_END))
+ address = (unsigned long) __va(phys_addr);
#endif
- if (!kernel_map || pte_present(pfn_pte(0, prot))) {
- err = __change_page_attr(address, pfn, prot);
- if (err)
- return err;
- }
+ err = __change_page_attr(address, cpa);
+ if (err)
+ return err;
#ifdef CONFIG_X86_64
/*
- * Handle kernel mapping too which aliases part of
- * lowmem:
+ * If the physical address is inside the kernel map, we need
+ * to touch the high mapped kernel as well:
*/
- if (__pa(address) < KERNEL_TEXT_SIZE) {
- unsigned long addr2;
- pgprot_t prot2;
-
- addr2 = __START_KERNEL_map + __pa(address);
- /* Make sure the kernel mappings stay executable */
- prot2 = pte_pgprot(pte_mkexec(pfn_pte(0, prot)));
- err = __change_page_attr(addr2, pfn, prot2);
+ if (within(phys_addr, 0, KERNEL_TEXT_SIZE)) {
+ /*
+ * Calc the high mapping address. See __phys_addr()
+ * for the non obvious details.
+ *
+ * Note that NX and other required permissions are
+ * checked in static_protections().
+ */
+ address = phys_addr + HIGH_MAP_START - phys_base;
+
+ /*
+ * Our high aliases are imprecise, because we check
+ * everything between 0 and KERNEL_TEXT_SIZE, so do
+ * not propagate lookup failures back to users:
+ */
+ __change_page_attr(address, cpa);
}
#endif
-
return err;
}
-/**
- * change_page_attr_set - Change page table attributes in the linear mapping.
- * @addr: Virtual address in linear mapping.
- * @numpages: Number of pages to change
- * @prot: Protection/caching type bits to set (PAGE_*)
- *
- * Returns 0 on success, otherwise a negated errno.
- *
- * This should be used when a page is mapped with a different caching policy
- * than write-back somewhere - some CPUs do not like it when mappings with
- * different caching policies exist. This changes the page attributes of the
- * in kernel linear mapping too.
- *
- * The caller needs to ensure that there are no conflicting mappings elsewhere
- * (e.g. in user space) * This function only deals with the kernel linear map.
- *
- * This function is different from change_page_attr() in that only selected bits
- * are impacted, all other bits remain as is.
- */
-static int __change_page_attr_set(unsigned long addr, int numpages,
- pgprot_t prot)
+static int __change_page_attr_set_clr(struct cpa_data *cpa)
{
- pgprot_t current_prot, new_prot;
- int level;
- pte_t *pte;
- int i, ret;
-
- for (i = 0; i < numpages ; i++) {
-
- pte = lookup_address(addr, &level);
- if (!pte)
- return -EINVAL;
-
- current_prot = pte_pgprot(*pte);
-
- pgprot_val(new_prot) =
- pgprot_val(current_prot) | pgprot_val(prot);
-
- ret = change_page_attr_addr(addr, new_prot);
+ int ret, numpages = cpa->numpages;
+
+ while (numpages) {
+ /*
+ * Store the remaining nr of pages for the large page
+ * preservation check.
+ */
+ cpa->numpages = numpages;
+ ret = change_page_attr_addr(cpa);
if (ret)
return ret;
- addr += PAGE_SIZE;
- }
+ /*
+ * Adjust the number of pages with the result of the
+ * CPA operation. Either a large page has been
+ * preserved or a single page update happened.
+ */
+ BUG_ON(cpa->numpages > numpages);
+ numpages -= cpa->numpages;
+ cpa->vaddr += cpa->numpages * PAGE_SIZE;
+ }
return 0;
}
-static int change_page_attr_set(unsigned long addr, int numpages, pgprot_t prot)
+static inline int cache_attr(pgprot_t attr)
{
- int ret = __change_page_attr_set(addr, numpages, prot);
-
- global_flush_tlb();
- return ret;
-
+ return pgprot_val(attr) &
+ (_PAGE_PAT | _PAGE_PAT_LARGE | _PAGE_PWT | _PAGE_PCD);
}
-/**
- * change_page_attr_clear - Change page table attributes in the linear mapping.
- * @addr: Virtual address in linear mapping.
- * @numpages: Number of pages to change
- * @prot: Protection/caching type bits to clear (PAGE_*)
- *
- * Returns 0 on success, otherwise a negated errno.
- *
- * This should be used when a page is mapped with a different caching policy
- * than write-back somewhere - some CPUs do not like it when mappings with
- * different caching policies exist. This changes the page attributes of the
- * in kernel linear mapping too.
- *
- * The caller needs to ensure that there are no conflicting mappings elsewhere
- * (e.g. in user space) * This function only deals with the kernel linear map.
- *
- * This function is different from change_page_attr() in that only selected bits
- * are impacted, all other bits remain as is.
- */
-static int __change_page_attr_clear(unsigned long addr, int numpages,
- pgprot_t prot)
+static int change_page_attr_set_clr(unsigned long addr, int numpages,
+ pgprot_t mask_set, pgprot_t mask_clr)
{
- pgprot_t current_prot, new_prot;
- int level;
- pte_t *pte;
- int i, ret;
+ struct cpa_data cpa;
+ int ret, cache;
- for (i = 0; i < numpages; i++) {
+ /*
+ * Check, if we are requested to change a not supported
+ * feature:
+ */
+ mask_set = canon_pgprot(mask_set);
+ mask_clr = canon_pgprot(mask_clr);
+ if (!pgprot_val(mask_set) && !pgprot_val(mask_clr))
+ return 0;
- pte = lookup_address(addr, &level);
- if (!pte)
- return -EINVAL;
+ cpa.vaddr = addr;
+ cpa.numpages = numpages;
+ cpa.mask_set = mask_set;
+ cpa.mask_clr = mask_clr;
+ cpa.flushtlb = 0;
- current_prot = pte_pgprot(*pte);
+ ret = __change_page_attr_set_clr(&cpa);
- pgprot_val(new_prot) =
- pgprot_val(current_prot) & ~pgprot_val(prot);
+ /*
+ * Check whether we really changed something:
+ */
+ if (!cpa.flushtlb)
+ return ret;
- ret = change_page_attr_addr(addr, new_prot);
- if (ret)
- return ret;
- addr += PAGE_SIZE;
- }
+ /*
+ * No need to flush, when we did not set any of the caching
+ * attributes:
+ */
+ cache = cache_attr(mask_set);
- return 0;
+ /*
+ * On success we use clflush, when the CPU supports it to
+ * avoid the wbindv. If the CPU does not support it and in the
+ * error case we fall back to cpa_flush_all (which uses
+ * wbindv):
+ */
+ if (!ret && cpu_has_clflush)
+ cpa_flush_range(addr, numpages, cache);
+ else
+ cpa_flush_all(cache);
+
+ return ret;
}
-static int change_page_attr_clear(unsigned long addr, int numpages,
- pgprot_t prot)
+static inline int change_page_attr_set(unsigned long addr, int numpages,
+ pgprot_t mask)
{
- int ret = __change_page_attr_clear(addr, numpages, prot);
-
- global_flush_tlb();
- return ret;
+ return change_page_attr_set_clr(addr, numpages, mask, __pgprot(0));
+}
+static inline int change_page_attr_clear(unsigned long addr, int numpages,
+ pgprot_t mask)
+{
+ return change_page_attr_set_clr(addr, numpages, __pgprot(0), mask);
}
int set_memory_uc(unsigned long addr, int numpages)
return set_memory_rw(addr, numpages);
}
-
#ifdef CONFIG_DEBUG_PAGEALLOC
static int __set_pages_p(struct page *page, int numpages)
{
- unsigned long addr = (unsigned long)page_address(page);
+ struct cpa_data cpa = { .vaddr = (unsigned long) page_address(page),
+ .numpages = numpages,
+ .mask_set = __pgprot(_PAGE_PRESENT | _PAGE_RW),
+ .mask_clr = __pgprot(0)};
- return __change_page_attr_set(addr, numpages,
- __pgprot(_PAGE_PRESENT | _PAGE_RW));
+ return __change_page_attr_set_clr(&cpa);
}
static int __set_pages_np(struct page *page, int numpages)
{
- unsigned long addr = (unsigned long)page_address(page);
+ struct cpa_data cpa = { .vaddr = (unsigned long) page_address(page),
+ .numpages = numpages,
+ .mask_set = __pgprot(0),
+ .mask_clr = __pgprot(_PAGE_PRESENT | _PAGE_RW)};
- return __change_page_attr_clear(addr, numpages,
- __pgprot(_PAGE_PRESENT));
+ return __change_page_attr_set_clr(&cpa);
}
void kernel_map_pages(struct page *page, int numpages, int enable)
projects / linux-2.6-omap-h63xx.git / blobdiff