xen/riscv: introduce setup_initial_pages

The idea was taken from xvisor but the following changes
were done:
* Use only a minimal part of the code enough to enable MMU
* rename {_}setup_initial_pagetables functions
* add an argument for setup_initial_mapping to have
  an opportunity to make set PTE flags.
* update setup_initial_pagetables function to map sections
  with correct PTE flags.
* Rewrite enable_mmu() to C.
* map linker addresses range to load addresses range without
  1:1 mapping. It will be 1:1 only in case when
  load_start_addr is equal to linker_start_addr.
* add safety checks such as:
  * Xen size is less than page size
  * linker addresses range doesn't overlap load addresses
    range
* Rework macros {THIRD,SECOND,FIRST,ZEROETH}_{SHIFT,MASK}
* change PTE_LEAF_DEFAULT to RW instead of RWX.
* Remove phys_offset as it is not used now
* Remove alignment  of {map, pa}_start &= XEN_PT_LEVEL_MAP_MASK(0);
  in  setup_inital_mapping() as they should be already aligned.
  Make a check that {map_pa}_start are aligned.
* Remove clear_pagetables() as initial pagetables will be
  zeroed during bss initialization
* Remove __attribute__((section(".entry")) for setup_initial_pagetables()
  as there is no such section in xen.lds.S
* Update the argument of pte_is_valid() to "const pte_t *p"
* Add check that Xen's load address is aligned at 4k boundary
* Refactor setup_initial_pagetables() so it is mapping linker
  address range to load address range. After setup needed
  permissions for specific section ( such as .text, .rodata, etc )
  otherwise RW permission will be set by default.
* Add function to check that requested SATP_MODE is supported

Origin: git@github.com:xvisor/xvisor.git 9be2fdd7
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Bobby Eshleman <bobbyeshleman@gmail.com>

xen/arch/riscv/Makefile

@@ -1,5 +1,6 @@
 obj-$(CONFIG_EARLY_PRINTK) += early_printk.o
 obj-y += entry.o
+obj-y += mm.o
 obj-$(CONFIG_RISCV_64) += riscv64/
 obj-y += sbi.o
 obj-y += setup.o

xen/arch/riscv/include/asm/config.h

@@ -75,12 +75,24 @@
   name:
 #endif
 
-#define XEN_VIRT_START  _AT(UL, 0x80200000)
+#ifdef CONFIG_RISCV_64
+#define XEN_VIRT_START 0xFFFFFFFFC0000000 /* (_AC(-1, UL) + 1 - GB(1)) */
+#else
+#error "RV32 isn't supported"
+#endif
 
 #define SMP_CACHE_BYTES (1 << 6)
 
 #define STACK_SIZE PAGE_SIZE
 
+#ifdef CONFIG_RISCV_64
+#define CONFIG_PAGING_LEVELS 3
+#define RV_STAGE1_MODE SATP_MODE_SV39
+#else
+#define CONFIG_PAGING_LEVELS 2
+#define RV_STAGE1_MODE SATP_MODE_SV32
+#endif
+
 #endif /* __RISCV_CONFIG_H__ */
 /*
  * Local variables:

xen/arch/riscv/include/asm/current.h

@@ -0,0 +1,11 @@
+#ifndef __ASM_CURRENT_H
+#define __ASM_CURRENT_H
+
+#define switch_stack_and_jump(stack, fn) do {               \
+    asm volatile (                                          \
+            "mv sp, %0\n"                                   \
+            "j " #fn :: "r" (stack), "X" (fn) : "memory" ); \
+    unreachable();                                          \
+} while ( false )
+
+#endif /* __ASM_CURRENT_H */

xen/arch/riscv/include/asm/mm.h

@@ -0,0 +1,14 @@
+#ifndef _ASM_RISCV_MM_H
+#define _ASM_RISCV_MM_H
+
+#include <asm/page-bits.h>
+
+#define pfn_to_paddr(pfn) ((paddr_t)(pfn) << PAGE_SHIFT)
+#define paddr_to_pfn(pa)  ((unsigned long)((pa) >> PAGE_SHIFT))
+
+void setup_initial_pagetables(void);
+
+void enable_mmu(void);
+void cont_after_mmu_is_enabled(void);
+
+#endif /* _ASM_RISCV_MM_H */

xen/arch/riscv/include/asm/page-bits.h

@@ -4,4 +4,14 @@
 #define PAGE_SHIFT              12 /* 4 KiB Pages */
 #define PADDR_BITS              56 /* 44-bit PPN */
 
+#ifdef CONFIG_RISCV_64
+#define PAGETABLE_ORDER         (9)
+#else /* CONFIG_RISCV_32 */
+#define PAGETABLE_ORDER         (10)
+#endif
+
+#define PAGETABLE_ENTRIES       (1 << PAGETABLE_ORDER)
+
+#define PTE_PPN_SHIFT           10
+
 #endif /* __RISCV_PAGE_BITS_H__ */

xen/arch/riscv/include/asm/page.h

@@ -0,0 +1,61 @@
+#ifndef _ASM_RISCV_PAGE_H
+#define _ASM_RISCV_PAGE_H
+
+#include <xen/const.h>
+#include <xen/types.h>
+
+#include <asm/mm.h>
+#include <asm/page-bits.h>
+
+#define VPN_MASK                    (PAGETABLE_ENTRIES - 1UL)
+
+#define XEN_PT_LEVEL_ORDER(lvl)     ((lvl) * PAGETABLE_ORDER)
+#define XEN_PT_LEVEL_SHIFT(lvl)     (XEN_PT_LEVEL_ORDER(lvl) + PAGE_SHIFT)
+#define XEN_PT_LEVEL_SIZE(lvl)      (_AT(paddr_t, 1) << XEN_PT_LEVEL_SHIFT(lvl))
+#define XEN_PT_LEVEL_MAP_MASK(lvl)  (~(XEN_PT_LEVEL_SIZE(lvl) - 1))
+#define XEN_PT_LEVEL_MASK(lvl)      (VPN_MASK << XEN_PT_LEVEL_SHIFT(lvl))
+
+#define PTE_VALID                   BIT(0, UL)
+#define PTE_READABLE                BIT(1, UL)
+#define PTE_WRITABLE                BIT(2, UL)
+#define PTE_EXECUTABLE              BIT(3, UL)
+#define PTE_USER                    BIT(4, UL)
+#define PTE_GLOBAL                  BIT(5, UL)
+#define PTE_ACCESSED                BIT(6, UL)
+#define PTE_DIRTY                   BIT(7, UL)
+#define PTE_RSW                     (BIT(8, UL) | BIT(9, UL))
+
+#define PTE_LEAF_DEFAULT            (PTE_VALID | PTE_READABLE | PTE_WRITABLE)
+#define PTE_TABLE                   (PTE_VALID)
+
+/* Calculate the offsets into the pagetables for a given VA */
+#define pt_linear_offset(lvl, va)   ((va) >> XEN_PT_LEVEL_SHIFT(lvl))
+
+#define pt_index(lvl, va) (pt_linear_offset((lvl), (va)) & VPN_MASK)
+
+/* Page Table entry */
+typedef struct {
+#ifdef CONFIG_RISCV_64
+    uint64_t pte;
+#else
+    uint32_t pte;
+#endif
+} pte_t;
+
+static inline pte_t paddr_to_pte(paddr_t paddr,
+                                 unsigned int permissions)
+{
+    return (pte_t) { .pte = (paddr_to_pfn(paddr) << PTE_PPN_SHIFT) | permissions };
+}
+
+static inline paddr_t pte_to_paddr(pte_t pte)
+{
+    return pfn_to_paddr(pte.pte >> PTE_PPN_SHIFT);
+}
+
+static inline bool pte_is_valid(pte_t p)
+{
+    return p.pte & PTE_VALID;
+}
+
+#endif /* _ASM_RISCV_PAGE_H */

xen/arch/riscv/include/asm/processor.h

@@ -69,6 +69,11 @@ static inline void die(void)
         wfi();
 }
 
+static inline void sfence_vma(void)
+{
+    asm volatile ( "sfence.vma" ::: "memory" );
+}
+
 #endif /* __ASSEMBLY__ */
 
 #endif /* _ASM_RISCV_PROCESSOR_H */

xen/arch/riscv/mm.c

@@ -0,0 +1,275 @@
+#include <xen/compiler.h>
+#include <xen/init.h>
+#include <xen/kernel.h>
+#include <xen/pfn.h>
+
+#include <asm/early_printk.h>
+#include <asm/csr.h>
+#include <asm/current.h>
+#include <asm/mm.h>
+#include <asm/page.h>
+#include <asm/processor.h>
+
+struct mmu_desc {
+    unsigned int num_levels;
+    unsigned int pgtbl_count;
+    pte_t *next_pgtbl;
+    pte_t *pgtbl_base;
+};
+
+extern unsigned char cpu0_boot_stack[STACK_SIZE];
+
+#define PHYS_OFFSET ((unsigned long)_start - XEN_VIRT_START)
+#define LOAD_TO_LINK(addr) ((addr) - PHYS_OFFSET)
+#define LINK_TO_LOAD(addr) ((addr) + PHYS_OFFSET)
+
+/*
+ * It is expected that Xen won't be more then 2 MB.
+ * The check in xen.lds.S guarantees that.
+ * At least 3 page tables (in case of Sv39 ) are needed to cover 2 MB.
+ * One for each page level table with PAGE_SIZE = 4 Kb.
+ *
+ * One L0 page table can cover 2 MB(512 entries of one page table * PAGE_SIZE).
+ *
+ * It might be needed one more page table in case when Xen load address
+ * isn't 2 MB aligned.
+ */
+#define PGTBL_INITIAL_COUNT ((CONFIG_PAGING_LEVELS - 1) + 1)
+
+pte_t __section(".bss.page_aligned") __aligned(PAGE_SIZE)
+stage1_pgtbl_root[PAGETABLE_ENTRIES];
+
+pte_t __section(".bss.page_aligned") __aligned(PAGE_SIZE)
+stage1_pgtbl_nonroot[PGTBL_INITIAL_COUNT * PAGETABLE_ENTRIES];
+
+#define HANDLE_PGTBL(curr_lvl_num)                                          \
+    index = pt_index(curr_lvl_num, page_addr);                              \
+    if ( pte_is_valid(pgtbl[index]) )                                       \
+    {                                                                       \
+        /* Find L{ 0-3 } table */                                           \
+        pgtbl = (pte_t *)pte_to_paddr(pgtbl[index]);                        \
+    }                                                                       \
+    else                                                                    \
+    {                                                                       \
+        /* Allocate new L{0-3} page table */                                \
+        if ( mmu_desc->pgtbl_count == PGTBL_INITIAL_COUNT )                 \
+        {                                                                   \
+            early_printk("(XEN) No initial table available\n");             \
+            /* panic(), BUG() or ASSERT() aren't ready now. */              \
+            die();                                                          \
+        }                                                                   \
+        mmu_desc->pgtbl_count++;                                            \
+        pgtbl[index] = paddr_to_pte((unsigned long)mmu_desc->next_pgtbl,    \
+                                    PTE_VALID);                             \
+        pgtbl = mmu_desc->next_pgtbl;                                       \
+        mmu_desc->next_pgtbl += PAGETABLE_ENTRIES;                          \
+    }
+
+static void __init setup_initial_mapping(struct mmu_desc *mmu_desc,
+                                         unsigned long map_start,
+                                         unsigned long map_end,
+                                         unsigned long pa_start)
+{
+    unsigned int index;
+    pte_t *pgtbl;
+    unsigned long page_addr;
+
+    if ( (unsigned long)_start % XEN_PT_LEVEL_SIZE(0) )
+    {
+        early_printk("(XEN) Xen should be loaded at 4k boundary\n");
+        die();
+    }
+
+    if ( (map_start & ~XEN_PT_LEVEL_MAP_MASK(0)) ||
+         (pa_start & ~XEN_PT_LEVEL_MAP_MASK(0)) )
+    {
+        early_printk("(XEN) map and pa start addresses should be aligned\n");
+        /* panic(), BUG() or ASSERT() aren't ready now. */
+        die();
+    }
+
+    for ( page_addr = map_start;
+          page_addr < map_end;
+          page_addr += XEN_PT_LEVEL_SIZE(0) )
+    {
+        pgtbl = mmu_desc->pgtbl_base;
+
+        switch ( mmu_desc->num_levels )
+        {
+        case 4: /* Level 3 */
+            HANDLE_PGTBL(3);
+        case 3: /* Level 2 */
+            HANDLE_PGTBL(2);
+        case 2: /* Level 1 */
+            HANDLE_PGTBL(1);
+        case 1: /* Level 0 */
+            {
+                unsigned long paddr = (page_addr - map_start) + pa_start;
+                unsigned int permissions = PTE_LEAF_DEFAULT;
+                pte_t pte_to_be_written;
+
+                index = pt_index(0, page_addr);
+
+                if ( is_kernel_text(LINK_TO_LOAD(page_addr)) ||
+                     is_kernel_inittext(LINK_TO_LOAD(page_addr)) )
+                    permissions =
+                        PTE_EXECUTABLE | PTE_READABLE | PTE_VALID;
+
+                if ( is_kernel_rodata(LINK_TO_LOAD(page_addr)) )
+                    permissions = PTE_READABLE | PTE_VALID;
+
+                pte_to_be_written = paddr_to_pte(paddr, permissions);
+
+                if ( !pte_is_valid(pgtbl[index]) )
+                    pgtbl[index] = pte_to_be_written;
+                else
+                {
+                    if ( (pgtbl[index].pte ^ pte_to_be_written.pte) &
+                         ~(PTE_DIRTY | PTE_ACCESSED) )
+                    {
+                        early_printk("PTE overridden has occurred\n");
+                        /* panic(), <asm/bug.h> aren't ready now. */
+                        die();
+                    }
+                }
+            }
+        }
+    }
+}
+#undef HANDLE_PGTBL
+
+static bool __init check_pgtbl_mode_support(struct mmu_desc *mmu_desc,
+                                            unsigned long load_start)
+{
+    bool is_mode_supported = false;
+    unsigned int index;
+    unsigned int page_table_level = (mmu_desc->num_levels - 1);
+    unsigned level_map_mask = XEN_PT_LEVEL_MAP_MASK(page_table_level);
+
+    unsigned long aligned_load_start = load_start & level_map_mask;
+    unsigned long aligned_page_size = XEN_PT_LEVEL_SIZE(page_table_level);
+    unsigned long xen_size = (unsigned long)(_end - start);
+
+    if ( (load_start + xen_size) > (aligned_load_start + aligned_page_size) )
+    {
+        early_printk("please place Xen to be in range of PAGE_SIZE "
+                     "where PAGE_SIZE is XEN_PT_LEVEL_SIZE( {L3 | L2 | L1} ) "
+                     "depending on expected SATP_MODE \n"
+                     "XEN_PT_LEVEL_SIZE is defined in <asm/page.h>\n");
+        die();
+    }
+
+    index = pt_index(page_table_level, aligned_load_start);
+    stage1_pgtbl_root[index] = paddr_to_pte(aligned_load_start,
+                                            PTE_LEAF_DEFAULT | PTE_EXECUTABLE);
+
+    sfence_vma();
+    csr_write(CSR_SATP,
+              PFN_DOWN((unsigned long)stage1_pgtbl_root) |
+              RV_STAGE1_MODE << SATP_MODE_SHIFT);
+
+    if ( (csr_read(CSR_SATP) >> SATP_MODE_SHIFT) == RV_STAGE1_MODE )
+        is_mode_supported = true;
+
+    csr_write(CSR_SATP, 0);
+
+    sfence_vma();
+
+    /* Clean MMU root page table */
+    stage1_pgtbl_root[index] = paddr_to_pte(0x0, 0x0);
+
+    return is_mode_supported;
+}
+
+/*
+ * setup_initial_pagetables:
+ *
+ * Build the page tables for Xen that map the following:
+ *  1. Calculate page table's level numbers.
+ *  2. Init mmu description structure.
+ *  3. Check that linker addresses range doesn't overlap
+ *     with load addresses range
+ *  4. Map all linker addresses and load addresses ( it shouldn't
+ *     be 1:1 mapped and will be 1:1 mapped only in case if
+ *     linker address is equal to load address ) with
+ *     RW permissions by default.
+ *  5. Setup proper PTE permissions for each section.
+ */
+void __init setup_initial_pagetables(void)
+{
+    struct mmu_desc mmu_desc = { CONFIG_PAGING_LEVELS, 0, NULL, NULL };
+
+    /*
+     * Access to _start, _end is always PC-relative thereby when access
+     * them we will get load adresses of start and end of Xen.
+     * To get linker addresses LOAD_TO_LINK() is required to use.
+     */
+    unsigned long load_start    = (unsigned long)_start;
+    unsigned long load_end      = (unsigned long)_end;
+    unsigned long linker_start  = LOAD_TO_LINK(load_start);
+    unsigned long linker_end    = LOAD_TO_LINK(load_end);
+
+    if ( (linker_start != load_start) &&
+         (linker_start <= load_end) && (load_start <= linker_end) )
+    {
+        early_printk("(XEN) linker and load address ranges overlap\n");
+        die();
+    }
+
+    if ( !check_pgtbl_mode_support(&mmu_desc, load_start) )
+    {
+        early_printk("requested MMU mode isn't supported by CPU\n"
+                     "Please choose different in <asm/config.h>\n");
+        die();
+    }
+
+    mmu_desc.pgtbl_base = stage1_pgtbl_root;
+    mmu_desc.next_pgtbl = stage1_pgtbl_nonroot;
+
+    setup_initial_mapping(&mmu_desc,
+                          linker_start,
+                          linker_end,
+                          load_start);
+}
+
+void __init noreturn noinline enable_mmu()
+{
+    /*
+     * Calculate a linker time address of the mmu_is_enabled
+     * label and update CSR_STVEC with it.
+     * MMU is configured in a way where linker addresses are mapped
+     * on load addresses so in a case when linker addresses are not equal
+     * to load addresses, after MMU is enabled, it will cause
+     * an exception and jump to linker time addresses.
+     * Otherwise if load addresses are equal to linker addresses the code
+     * after mmu_is_enabled label will be executed without exception.
+     */
+    csr_write(CSR_STVEC, LOAD_TO_LINK((unsigned long)&&mmu_is_enabled));
+
+    /* Ensure page table writes precede loading the SATP */
+    sfence_vma();
+
+    /* Enable the MMU and load the new pagetable for Xen */
+    csr_write(CSR_SATP,
+              PFN_DOWN((unsigned long)stage1_pgtbl_root) |
+              RV_STAGE1_MODE << SATP_MODE_SHIFT);
+
+    asm volatile ( ".p2align 2" );
+ mmu_is_enabled:
+    /*
+     * Stack should be re-inited as:
+     * 1. Right now an address of the stack is relative to load time
+     *    addresses what will cause an issue in case of load start address
+     *    isn't equal to linker start address.
+     * 2. Addresses in stack are all load time relative which can be an
+     *    issue in case when load start address isn't equal to linker
+     *    start address.
+     *
+     * We can't return to the caller because the stack was reseted
+     * and it may have stash some variable on the stack.
+     * Jump to a brand new function as the stack was reseted
+     */
+
+    switch_stack_and_jump((unsigned long)cpu0_boot_stack + STACK_SIZE,
+                          cont_after_mmu_is_enabled);
+}

xen/arch/riscv/setup.c

@@ -2,6 +2,7 @@
 #include <xen/init.h>
 
 #include <asm/early_printk.h>
+#include <asm/mm.h>
 
 /* Xen stack for bringing up the first CPU. */
 unsigned char __initdata cpu0_boot_stack[STACK_SIZE]
@@ -26,3 +27,13 @@ void __init noreturn start_xen(unsigned long bootcpu_id,
 
     unreachable();
 }
+
+void __init noreturn cont_after_mmu_is_enabled(void)
+{
+    early_printk("All set up\n");
+
+    for ( ;; )
+        asm volatile ("wfi");
+
+    unreachable();
+}

xen/arch/riscv/xen.lds.S

@@ -172,3 +172,5 @@ ASSERT(IS_ALIGNED(__bss_end,        POINTER_ALIGN), "__bss_end is misaligned")
 
 ASSERT(!SIZEOF(.got),      ".got non-empty")
 ASSERT(!SIZEOF(.got.plt),  ".got.plt non-empty")
+
+ASSERT(_end - _start <= MB(2), "Xen too large for early-boot assumptions")