[PATCH] xtensa: Architecture support for Tensilica Xtensa Part 6
The attached patches provides part 6 of an architecture implementation for the
Tensilica Xtensa CPU series.
Signed-off-by: Chris Zankel <chris@zankel.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
diff --git a/include/asm-xtensa/pgtable.h b/include/asm-xtensa/pgtable.h
new file mode 100644
index 0000000..0bb6416
--- /dev/null
+++ b/include/asm-xtensa/pgtable.h
@@ -0,0 +1,468 @@
+/*
+ * linux/include/asm-xtensa/page.h
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version2 as
+ * published by the Free Software Foundation.
+ *
+ * Copyright (C) 2001 - 2005 Tensilica Inc.
+ */
+
+#ifndef _XTENSA_PGTABLE_H
+#define _XTENSA_PGTABLE_H
+
+#include <asm-generic/pgtable-nopmd.h>
+#include <asm/page.h>
+
+/* Assertions. */
+
+#ifdef CONFIG_MMU
+
+
+#if (XCHAL_MMU_RINGS < 2)
+# error Linux build assumes at least 2 ring levels.
+#endif
+
+#if (XCHAL_MMU_CA_BITS != 4)
+# error We assume exactly four bits for CA.
+#endif
+
+#if (XCHAL_MMU_SR_BITS != 0)
+# error We have no room for SR bits.
+#endif
+
+/*
+ * Use the first min-wired way for mapping page-table pages.
+ * Page coloring requires a second min-wired way.
+ */
+
+#if (XCHAL_DTLB_MINWIRED_SETS == 0)
+# error Need a min-wired way for mapping page-table pages
+#endif
+
+#define DTLB_WAY_PGTABLE XCHAL_DTLB_SET(XCHAL_DTLB_MINWIRED_SET0, WAY)
+
+#if (DCACHE_WAY_SIZE > PAGE_SIZE) && XCHAL_DCACHE_IS_WRITEBACK
+# if XCHAL_DTLB_SET(XCHAL_DTLB_MINWIRED_SET0, WAYS) >= 2
+# define DTLB_WAY_DCACHE_ALIAS0 (DTLB_WAY_PGTABLE + 1)
+# define DTLB_WAY_DCACHE_ALIAS1 (DTLB_WAY_PGTABLE + 2)
+# else
+# error Page coloring requires its own wired dtlb way!
+# endif
+#endif
+
+#endif /* CONFIG_MMU */
+
+/*
+ * We only use two ring levels, user and kernel space.
+ */
+
+#define USER_RING 1 /* user ring level */
+#define KERNEL_RING 0 /* kernel ring level */
+
+/*
+ * The Xtensa architecture port of Linux has a two-level page table system,
+ * i.e. the logical three-level Linux page table layout are folded.
+ * Each task has the following memory page tables:
+ *
+ * PGD table (page directory), ie. 3rd-level page table:
+ * One page (4 kB) of 1024 (PTRS_PER_PGD) pointers to PTE tables
+ * (Architectures that don't have the PMD folded point to the PMD tables)
+ *
+ * The pointer to the PGD table for a given task can be retrieved from
+ * the task structure (struct task_struct*) t, e.g. current():
+ * (t->mm ? t->mm : t->active_mm)->pgd
+ *
+ * PMD tables (page middle-directory), ie. 2nd-level page tables:
+ * Absent for the Xtensa architecture (folded, PTRS_PER_PMD == 1).
+ *
+ * PTE tables (page table entry), ie. 1st-level page tables:
+ * One page (4 kB) of 1024 (PTRS_PER_PTE) PTEs with a special PTE
+ * invalid_pte_table for absent mappings.
+ *
+ * The individual pages are 4 kB big with special pages for the empty_zero_page.
+ */
+#define PGDIR_SHIFT 22
+#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
+#define PGDIR_MASK (~(PGDIR_SIZE-1))
+
+/*
+ * Entries per page directory level: we use two-level, so
+ * we don't really have any PMD directory physically.
+ */
+#define PTRS_PER_PTE 1024
+#define PTRS_PER_PTE_SHIFT 10
+#define PTRS_PER_PMD 1
+#define PTRS_PER_PGD 1024
+#define PGD_ORDER 0
+#define PMD_ORDER 0
+#define USER_PTRS_PER_PGD (TASK_SIZE/PGDIR_SIZE)
+#define FIRST_USER_ADDRESS XCHAL_SEG_MAPPABLE_VADDR
+#define FIRST_USER_PGD_NR (FIRST_USER_ADDRESS >> PGDIR_SHIFT)
+
+/* virtual memory area. We keep a distance to other memory regions to be
+ * on the safe side. We also use this area for cache aliasing.
+ */
+
+// FIXME: virtual memory area must be configuration-dependent
+
+#define VMALLOC_START 0xC0000000
+#define VMALLOC_END 0xC7FF0000
+
+/* Xtensa Linux config PTE layout (when present):
+ * 31-12: PPN
+ * 11-6: Software
+ * 5-4: RING
+ * 3-0: CA
+ *
+ * Similar to the Alpha and MIPS ports, we need to keep track of the ref
+ * and mod bits in software. We have a software "you can read
+ * from this page" bit, and a hardware one which actually lets the
+ * process read from the page. On the same token we have a software
+ * writable bit and the real hardware one which actually lets the
+ * process write to the page.
+ *
+ * See further below for PTE layout for swapped-out pages.
+ */
+
+#define _PAGE_VALID (1<<0) /* hardware: page is accessible */
+#define _PAGE_WRENABLE (1<<1) /* hardware: page is writable */
+
+/* None of these cache modes include MP coherency: */
+#define _PAGE_NO_CACHE (0<<2) /* bypass, non-speculative */
+#if XCHAL_DCACHE_IS_WRITEBACK
+# define _PAGE_WRITEBACK (1<<2) /* write back */
+# define _PAGE_WRITETHRU (2<<2) /* write through */
+#else
+# define _PAGE_WRITEBACK (1<<2) /* assume write through */
+# define _PAGE_WRITETHRU (1<<2)
+#endif
+#define _PAGE_NOALLOC (3<<2) /* don't allocate cache,if not cached */
+#define _CACHE_MASK (3<<2)
+
+#define _PAGE_USER (1<<4) /* user access (ring=1) */
+#define _PAGE_KERNEL (0<<4) /* kernel access (ring=0) */
+
+/* Software */
+#define _PAGE_RW (1<<6) /* software: page writable */
+#define _PAGE_DIRTY (1<<7) /* software: page dirty */
+#define _PAGE_ACCESSED (1<<8) /* software: page accessed (read) */
+#define _PAGE_FILE (1<<9) /* nonlinear file mapping*/
+
+#define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_ACCESSED | _CACHE_MASK | _PAGE_DIRTY)
+#define _PAGE_PRESENT ( _PAGE_VALID | _PAGE_WRITEBACK | _PAGE_ACCESSED)
+
+#ifdef CONFIG_MMU
+
+# define PAGE_NONE __pgprot(_PAGE_PRESENT)
+# define PAGE_SHARED __pgprot(_PAGE_PRESENT | _PAGE_USER | _PAGE_RW)
+# define PAGE_COPY __pgprot(_PAGE_PRESENT | _PAGE_USER)
+# define PAGE_READONLY __pgprot(_PAGE_PRESENT | _PAGE_USER)
+# define PAGE_KERNEL __pgprot(_PAGE_PRESENT | _PAGE_KERNEL | _PAGE_WRENABLE)
+# define PAGE_INVALID __pgprot(_PAGE_USER)
+
+# if (DCACHE_WAY_SIZE > PAGE_SIZE)
+# define PAGE_DIRECTORY __pgprot(_PAGE_VALID | _PAGE_ACCESSED | _PAGE_KERNEL)
+# else
+# define PAGE_DIRECTORY __pgprot(_PAGE_PRESENT | _PAGE_KERNEL)
+# endif
+
+#else /* no mmu */
+
+# define PAGE_NONE __pgprot(0)
+# define PAGE_SHARED __pgprot(0)
+# define PAGE_COPY __pgprot(0)
+# define PAGE_READONLY __pgprot(0)
+# define PAGE_KERNEL __pgprot(0)
+
+#endif
+
+/*
+ * On certain configurations of Xtensa MMUs (eg. the initial Linux config),
+ * the MMU can't do page protection for execute, and considers that the same as
+ * read. Also, write permissions may imply read permissions.
+ * What follows is the closest we can get by reasonable means..
+ * See linux/mm/mmap.c for protection_map[] array that uses these definitions.
+ */
+#define __P000 PAGE_NONE /* private --- */
+#define __P001 PAGE_READONLY /* private --r */
+#define __P010 PAGE_COPY /* private -w- */
+#define __P011 PAGE_COPY /* private -wr */
+#define __P100 PAGE_READONLY /* private x-- */
+#define __P101 PAGE_READONLY /* private x-r */
+#define __P110 PAGE_COPY /* private xw- */
+#define __P111 PAGE_COPY /* private xwr */
+
+#define __S000 PAGE_NONE /* shared --- */
+#define __S001 PAGE_READONLY /* shared --r */
+#define __S010 PAGE_SHARED /* shared -w- */
+#define __S011 PAGE_SHARED /* shared -wr */
+#define __S100 PAGE_READONLY /* shared x-- */
+#define __S101 PAGE_READONLY /* shared x-r */
+#define __S110 PAGE_SHARED /* shared xw- */
+#define __S111 PAGE_SHARED /* shared xwr */
+
+#ifndef __ASSEMBLY__
+
+#define pte_ERROR(e) \
+ printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e))
+#define pgd_ERROR(e) \
+ printk("%s:%d: bad pgd entry %08lx.\n", __FILE__, __LINE__, pgd_val(e))
+
+extern unsigned long empty_zero_page[1024];
+
+#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
+
+extern pgd_t swapper_pg_dir[PAGE_SIZE/sizeof(pgd_t)];
+
+/*
+ * The pmd contains the kernel virtual address of the pte page.
+ */
+#define pmd_page_kernel(pmd) ((unsigned long)(pmd_val(pmd) & PAGE_MASK))
+#define pmd_page(pmd) virt_to_page(pmd_val(pmd))
+
+/*
+ * The following only work if pte_present() is true.
+ */
+#define pte_none(pte) (!(pte_val(pte) ^ _PAGE_USER))
+#define pte_present(pte) (pte_val(pte) & _PAGE_VALID)
+#define pte_clear(mm,addr,ptep) \
+ do { update_pte(ptep, __pte(_PAGE_USER)); } while(0)
+
+#define pmd_none(pmd) (!pmd_val(pmd))
+#define pmd_present(pmd) (pmd_val(pmd) & PAGE_MASK)
+#define pmd_clear(pmdp) do { set_pmd(pmdp, __pmd(0)); } while (0)
+#define pmd_bad(pmd) (pmd_val(pmd) & ~PAGE_MASK)
+
+/* Note: We use the _PAGE_USER bit to indicate write-protect kernel memory */
+
+static inline int pte_read(pte_t pte) { return pte_val(pte) & _PAGE_USER; }
+static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW; }
+static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY; }
+static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED; }
+static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE; }
+static inline pte_t pte_wrprotect(pte_t pte) { pte_val(pte) &= ~(_PAGE_RW | _PAGE_WRENABLE); return pte; }
+static inline pte_t pte_rdprotect(pte_t pte) { pte_val(pte) &= ~_PAGE_USER; return pte; }
+static inline pte_t pte_mkclean(pte_t pte) { pte_val(pte) &= ~_PAGE_DIRTY; return pte; }
+static inline pte_t pte_mkold(pte_t pte) { pte_val(pte) &= ~_PAGE_ACCESSED; return pte; }
+static inline pte_t pte_mkread(pte_t pte) { pte_val(pte) |= _PAGE_USER; return pte; }
+static inline pte_t pte_mkdirty(pte_t pte) { pte_val(pte) |= _PAGE_DIRTY; return pte; }
+static inline pte_t pte_mkyoung(pte_t pte) { pte_val(pte) |= _PAGE_ACCESSED; return pte; }
+static inline pte_t pte_mkwrite(pte_t pte) { pte_val(pte) |= _PAGE_RW; return pte; }
+
+/*
+ * Conversion functions: convert a page and protection to a page entry,
+ * and a page entry and page directory to the page they refer to.
+ */
+#define pte_pfn(pte) (pte_val(pte) >> PAGE_SHIFT)
+#define pte_same(a,b) (pte_val(a) == pte_val(b))
+#define pte_page(x) pfn_to_page(pte_pfn(x))
+#define pfn_pte(pfn, prot) __pte(((pfn) << PAGE_SHIFT) | pgprot_val(prot))
+#define mk_pte(page, prot) pfn_pte(page_to_pfn(page), prot)
+
+extern inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
+{
+ return __pte((pte_val(pte) & _PAGE_CHG_MASK) | pgprot_val(newprot));
+}
+
+/*
+ * Certain architectures need to do special things when pte's
+ * within a page table are directly modified. Thus, the following
+ * hook is made available.
+ */
+static inline void update_pte(pte_t *ptep, pte_t pteval)
+{
+ *ptep = pteval;
+#if (DCACHE_WAY_SIZE > PAGE_SIZE) && XCHAL_DCACHE_IS_WRITEBACK
+ __asm__ __volatile__ ("memw; dhwb %0, 0; dsync" :: "a" (ptep));
+#endif
+}
+
+extern inline void
+set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pteval)
+{
+ update_pte(ptep, pteval);
+}
+
+
+extern inline void
+set_pmd(pmd_t *pmdp, pmd_t pmdval)
+{
+ *pmdp = pmdval;
+#if (DCACHE_WAY_SIZE > PAGE_SIZE) && XCHAL_DCACHE_IS_WRITEBACK
+ __asm__ __volatile__ ("memw; dhwb %0, 0; dsync" :: "a" (pmdp));
+#endif
+}
+
+
+static inline int
+ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr,
+ pte_t *ptep)
+{
+ pte_t pte = *ptep;
+ if (!pte_young(pte))
+ return 0;
+ update_pte(ptep, pte_mkold(pte));
+ return 1;
+}
+
+static inline int
+ptep_test_and_clear_dirty(struct vm_area_struct *vma, unsigned long addr,
+ pte_t *ptep)
+{
+ pte_t pte = *ptep;
+ if (!pte_dirty(pte))
+ return 0;
+ update_pte(ptep, pte_mkclean(pte));
+ return 1;
+}
+
+static inline pte_t
+ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+{
+ pte_t pte = *ptep;
+ pte_clear(mm, addr, ptep);
+ return pte;
+}
+
+static inline void
+ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+{
+ pte_t pte = *ptep;
+ update_pte(ptep, pte_wrprotect(pte));
+}
+
+/* to find an entry in a kernel page-table-directory */
+#define pgd_offset_k(address) pgd_offset(&init_mm, address)
+
+/* to find an entry in a page-table-directory */
+#define pgd_offset(mm,address) ((mm)->pgd + pgd_index(address))
+
+#define pgd_index(address) ((address) >> PGDIR_SHIFT)
+
+/* Find an entry in the second-level page table.. */
+#define pmd_offset(dir,address) ((pmd_t*)(dir))
+
+/* Find an entry in the third-level page table.. */
+#define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
+#define pte_offset_kernel(dir,addr) \
+ ((pte_t*) pmd_page_kernel(*(dir)) + pte_index(addr))
+#define pte_offset_map(dir,addr) pte_offset_kernel((dir),(addr))
+#define pte_offset_map_nested(dir,addr) pte_offset_kernel((dir),(addr))
+
+#define pte_unmap(pte) do { } while (0)
+#define pte_unmap_nested(pte) do { } while (0)
+
+
+/*
+ * Encode and decode a swap entry.
+ * Each PTE in a process VM's page table is either:
+ * "present" -- valid and not swapped out, protection bits are meaningful;
+ * "not present" -- which further subdivides in these two cases:
+ * "none" -- no mapping at all; identified by pte_none(), set by pte_clear(
+ * "swapped out" -- the page is swapped out, and the SWP macros below
+ * are used to store swap file info in the PTE itself.
+ *
+ * In the Xtensa processor MMU, any PTE entries in user space (or anywhere
+ * in virtual memory that can map differently across address spaces)
+ * must have a correct ring value that represents the RASID field that
+ * is changed when switching address spaces. Eg. such PTE entries cannot
+ * be set to ring zero, because that can cause a (global) kernel ASID
+ * entry to be created in the TLBs (even with invalid cache attribute),
+ * potentially causing a multihit exception when going back to another
+ * address space that mapped the same virtual address at another ring.
+ *
+ * SO: we avoid using ring bits (_PAGE_RING_MASK) in "not present" PTEs.
+ * We also avoid using the _PAGE_VALID bit which must be zero for non-present
+ * pages.
+ *
+ * We end up with the following available bits: 1..3 and 7..31.
+ * We don't bother with 1..3 for now (we can use them later if needed),
+ * and chose to allocate 6 bits for SWP_TYPE and the remaining 19 bits
+ * for SWP_OFFSET. At least 5 bits are needed for SWP_TYPE, because it
+ * is currently implemented as an index into swap_info[MAX_SWAPFILES]
+ * and MAX_SWAPFILES is currently defined as 32 in <linux/swap.h>.
+ * However, for some reason all other architectures in the 2.4 kernel
+ * reserve either 6, 7, or 8 bits so I'll not detract from that for now. :)
+ * SWP_OFFSET is an offset into the swap file in page-size units, so
+ * with 4 kB pages, 19 bits supports a maximum swap file size of 2 GB.
+ *
+ * FIXME: 2 GB isn't very big. Other bits can be used to allow
+ * larger swap sizes. In the meantime, it appears relatively easy to get
+ * around the 2 GB limitation by simply using multiple swap files.
+ */
+
+#define __swp_type(entry) (((entry).val >> 7) & 0x3f)
+#define __swp_offset(entry) ((entry).val >> 13)
+#define __swp_entry(type,offs) ((swp_entry_t) {((type) << 7) | ((offs) << 13)})
+#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
+#define __swp_entry_to_pte(x) ((pte_t) { (x).val })
+
+#define PTE_FILE_MAX_BITS 29
+#define pte_to_pgoff(pte) (pte_val(pte) >> 3)
+#define pgoff_to_pte(off) ((pte_t) { ((off) << 3) | _PAGE_FILE })
+
+
+#endif /* !defined (__ASSEMBLY__) */
+
+
+#ifdef __ASSEMBLY__
+
+/* Assembly macro _PGD_INDEX is the same as C pgd_index(unsigned long),
+ * _PGD_OFFSET as C pgd_offset(struct mm_struct*, unsigned long),
+ * _PMD_OFFSET as C pmd_offset(pgd_t*, unsigned long)
+ * _PTE_OFFSET as C pte_offset(pmd_t*, unsigned long)
+ *
+ * Note: We require an additional temporary register which can be the same as
+ * the register that holds the address.
+ *
+ * ((pte_t*) ((unsigned long)(pmd_val(*pmd) & PAGE_MASK)) + pte_index(addr))
+ *
+ */
+#define _PGD_INDEX(rt,rs) extui rt, rs, PGDIR_SHIFT, 32-PGDIR_SHIFT
+#define _PTE_INDEX(rt,rs) extui rt, rs, PAGE_SHIFT, PTRS_PER_PTE_SHIFT
+
+#define _PGD_OFFSET(mm,adr,tmp) l32i mm, mm, MM_PGD; \
+ _PGD_INDEX(tmp, adr); \
+ addx4 mm, tmp, mm
+
+#define _PTE_OFFSET(pmd,adr,tmp) _PTE_INDEX(tmp, adr); \
+ srli pmd, pmd, PAGE_SHIFT; \
+ slli pmd, pmd, PAGE_SHIFT; \
+ addx4 pmd, tmp, pmd
+
+#else
+
+extern void paging_init(void);
+
+#define kern_addr_valid(addr) (1)
+
+extern void update_mmu_cache(struct vm_area_struct * vma,
+ unsigned long address, pte_t pte);
+
+/*
+ * remap a physical address `phys' of size `size' with page protection `prot'
+ * into virtual address `from'
+ */
+#define io_remap_page_range(vma,from,phys,size,prot) \
+ remap_pfn_range(vma, from, (phys) >> PAGE_SHIFT, size, prot)
+
+
+/* No page table caches to init */
+
+#define pgtable_cache_init() do { } while (0)
+
+typedef pte_t *pte_addr_t;
+
+#endif /* !defined (__ASSEMBLY__) */
+
+#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
+#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
+#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
+#define __HAVE_ARCH_PTEP_SET_WRPROTECT
+#define __HAVE_ARCH_PTEP_MKDIRTY
+#define __HAVE_ARCH_PTE_SAME
+
+#include <asm-generic/pgtable.h>
+
+#endif /* _XTENSA_PGTABLE_H */