[PATCH v2 00/10] xen: arm: direct 1:1 map on 64 bit

[PATCH v2 00/10] xen: arm: direct 1:1 map on 64 bit

[Ian Campbell]

A pretty substantial portion of this is acked but the big patch "xen:
arm: Use a direct mapping of RAM on arm64" remains outstanding.

The main concern seems to be whether to ifdef vs moving code under
arm32/ arm64/. Since the majority of the code remains common[0] in this
case I prefer to use ifdefs since it keeps all the interrelated code in
a single place and avoids having to export internals like the static
page table arrays outside of this function and expose other functions
which we would rather the rest of the hypervisor didn't get at.

Other patches which need acks are:

xen: arm: Add zeroeth level page table macros and defines

        I think this one just got missed, it's pretty trivial IMHO.
        
xen: arm: define a macro to get this CPUs page table root

        Stefano didn't see the point but it doesn't really become clear
        until a later patch. I've expanded the changelog.
        
[0] Out of 1255 lines in mm.c there are 45 lines of 64 bit only code and
64 lines of 32 bit only code.

[PATCH v2 01/10] xen: arm: Use _AC macros in config.h

[Ian Campbell]

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
 xen/include/asm-arm/config.h |   18 ++++++++++--------
 1 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/xen/include/asm-arm/config.h b/xen/include/asm-arm/config.h
index e3cfaf1..7e5daa0 100644
--- a/xen/include/asm-arm/config.h
+++ b/xen/include/asm-arm/config.h
@@ -72,6 +72,8 @@
   END(name)
 #endif
 
+#include <xen/const.h>
+
 /*
  * Memory layout:
  *  0  -   2M   Unmapped
@@ -91,14 +93,14 @@
  *   - in setup_pagetables() when relocating Xen.
  */
 
-#define XEN_VIRT_START         mk_unsigned_long(0x00200000)
-#define FIXMAP_ADDR(n)        (mk_unsigned_long(0x00400000) + (n) * PAGE_SIZE)
-#define BOOT_MISC_VIRT_START   mk_unsigned_long(0x00600000)
-#define FRAMETABLE_VIRT_START  mk_unsigned_long(0x02000000)
-#define VMAP_VIRT_START        mk_unsigned_long(0x10000000)
-#define XENHEAP_VIRT_START     mk_unsigned_long(0x40000000)
-#define DOMHEAP_VIRT_START     mk_unsigned_long(0x80000000)
-#define DOMHEAP_VIRT_END       mk_unsigned_long(0xffffffff)
+#define XEN_VIRT_START         _AC(0x00200000,UL)
+#define FIXMAP_ADDR(n)        (_AC(0x00400000,UL) + (n) * PAGE_SIZE)
+#define BOOT_MISC_VIRT_START   _AC(0x00600000,UL)
+#define FRAMETABLE_VIRT_START  _AC(0x02000000,UL)
+#define VMAP_VIRT_START        _AC(0x10000000,UL)
+#define XENHEAP_VIRT_START     _AC(0x40000000,UL)
+#define DOMHEAP_VIRT_START     _AC(0x80000000,UL)
+#define DOMHEAP_VIRT_END       _AC(0xffffffff,UL)
 
 #define VMAP_VIRT_END          XENHEAP_VIRT_START
 #define HYPERVISOR_VIRT_START  XEN_VIRT_START
-- 
1.7.2.5

[PATCH v2 02/10] xen: arm: Add zeroeth level page table macros and defines

[Ian Campbell]

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
---
 xen/include/asm-arm/page.h |   10 +++++++++-
 1 files changed, 9 insertions(+), 1 deletions(-)

diff --git a/xen/include/asm-arm/page.h b/xen/include/asm-arm/page.h
index 41e9eff..6317090 100644
--- a/xen/include/asm-arm/page.h
+++ b/xen/include/asm-arm/page.h
@@ -337,7 +337,15 @@ static inline int gva_to_ipa(vaddr_t va, paddr_t *paddr)
 #define second_table_offset(va) TABLE_OFFSET(second_linear_offset(va))
 #define third_table_offset(va)  TABLE_OFFSET(third_linear_offset(va))
 
-#define clear_page(page)memset((void *)(page), 0, PAGE_SIZE)
+#ifdef CONFIG_ARM_64
+#define ZEROETH_SHIFT  (FIRST_SHIFT + LPAE_SHIFT)
+#define ZEROETH_SIZE   (1u << ZEROETH_SHIFT)
+#define ZEROETH_MASK   (~(ZEROETH_SIZE - 1))
+#define zeroeth_linear_offset(va) ((va) >> ZEROETH_SHIFT)
+#define zeroeth_table_offset(va)  TABLE_OFFSET(zeroeth_linear_offset(va))
+#endif
+
+#define clear_page(page) memset((void *)(page), 0, PAGE_SIZE)
 
 #define PAGE_ALIGN(x) (((x) + PAGE_SIZE - 1) & PAGE_MASK)
 
-- 
1.7.2.5

Re: [PATCH v2 02/10] xen: arm: Add zeroeth level page table macros and defines

[Tim Deegan]

At 14:31 +0100 on 07 Aug (1375885891), Ian Campbell wrote:
> --- a/xen/include/asm-arm/page.h
> +++ b/xen/include/asm-arm/page.h
> @@ -337,7 +337,15 @@ static inline int gva_to_ipa(vaddr_t va, paddr_t *paddr)
>  #define second_table_offset(va) TABLE_OFFSET(second_linear_offset(va))
>  #define third_table_offset(va)  TABLE_OFFSET(third_linear_offset(va))
>  
> -#define clear_page(page)memset((void *)(page), 0, PAGE_SIZE)
> +#ifdef CONFIG_ARM_64
> +#define ZEROETH_SHIFT  (FIRST_SHIFT + LPAE_SHIFT)
> +#define ZEROETH_SIZE   (1u << ZEROETH_SHIFT)
> +#define ZEROETH_MASK   (~(ZEROETH_SIZE - 1))
> +#define zeroeth_linear_offset(va) ((va) >> ZEROETH_SHIFT)
> +#define zeroeth_table_offset(va)  TABLE_OFFSET(zeroeth_linear_offset(va))
> +#endif

I'd prefer to see these defined alongside the other levels.  There's no
harm in defining them on 32-bit builds.  Also, please update the comment
above LPAE_SHIFT to describe the 4-level alternative. 

Cheers,

Tim.

Re: [PATCH v2 02/10] xen: arm: Add zeroeth level page table macros and defines

[Ian Campbell]

On Thu, 2013-08-08 at 11:09 +0100, Tim Deegan wrote:
> At 14:31 +0100 on 07 Aug (1375885891), Ian Campbell wrote:
> > --- a/xen/include/asm-arm/page.h
> > +++ b/xen/include/asm-arm/page.h
> > @@ -337,7 +337,15 @@ static inline int gva_to_ipa(vaddr_t va, paddr_t *paddr)
> >  #define second_table_offset(va) TABLE_OFFSET(second_linear_offset(va))
> >  #define third_table_offset(va)  TABLE_OFFSET(third_linear_offset(va))
> >  
> > -#define clear_page(page)memset((void *)(page), 0, PAGE_SIZE)
> > +#ifdef CONFIG_ARM_64
> > +#define ZEROETH_SHIFT  (FIRST_SHIFT + LPAE_SHIFT)
> > +#define ZEROETH_SIZE   (1u << ZEROETH_SHIFT)
> > +#define ZEROETH_MASK   (~(ZEROETH_SIZE - 1))
> > +#define zeroeth_linear_offset(va) ((va) >> ZEROETH_SHIFT)
> > +#define zeroeth_table_offset(va)  TABLE_OFFSET(zeroeth_linear_offset(va))
> > +#endif
> 
> I'd prefer to see these defined alongside the other levels.  There's no
> harm in defining them on 32-bit builds.

My worry was that they would get used on 32-bit code paths by mistake,
which is a build error with this version but a (possibly obscure)
runtime error with the alternative.

> Also, please update the comment above LPAE_SHIFT to describe the 4-level alternative. 

Will do.

Re: [PATCH v2 02/10] xen: arm: Add zeroeth level page table macros and defines

[Tim Deegan]

At 11:12 +0100 on 08 Aug (1375960367), Ian Campbell wrote:
> On Thu, 2013-08-08 at 11:09 +0100, Tim Deegan wrote:
> > At 14:31 +0100 on 07 Aug (1375885891), Ian Campbell wrote:
> > > --- a/xen/include/asm-arm/page.h
> > > +++ b/xen/include/asm-arm/page.h
> > > @@ -337,7 +337,15 @@ static inline int gva_to_ipa(vaddr_t va, paddr_t *paddr)
> > >  #define second_table_offset(va) TABLE_OFFSET(second_linear_offset(va))
> > >  #define third_table_offset(va)  TABLE_OFFSET(third_linear_offset(va))
> > >  
> > > -#define clear_page(page)memset((void *)(page), 0, PAGE_SIZE)
> > > +#ifdef CONFIG_ARM_64
> > > +#define ZEROETH_SHIFT  (FIRST_SHIFT + LPAE_SHIFT)
> > > +#define ZEROETH_SIZE   (1u << ZEROETH_SHIFT)
> > > +#define ZEROETH_MASK   (~(ZEROETH_SIZE - 1))
> > > +#define zeroeth_linear_offset(va) ((va) >> ZEROETH_SHIFT)
> > > +#define zeroeth_table_offset(va)  TABLE_OFFSET(zeroeth_linear_offset(va))
> > > +#endif
> > 
> > I'd prefer to see these defined alongside the other levels.  There's no
> > harm in defining them on 32-bit builds.
> 
> My worry was that they would get used on 32-bit code paths by mistake,
> which is a build error with this version but a (possibly obscure)
> runtime error with the alternative.

Isn't at a build error to shift 32-bit values by 39 bits?  I guess is
more reliable, though; stet.

Tim.

> > Also, please update the comment above LPAE_SHIFT to describe the 4-level alternative. 
> 
> Will do.
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

Re: [PATCH v2 02/10] xen: arm: Add zeroeth level page table macros and defines

[Ian Campbell]

On Thu, 2013-08-08 at 12:08 +0100, Tim Deegan wrote:
> At 11:12 +0100 on 08 Aug (1375960367), Ian Campbell wrote:
> > On Thu, 2013-08-08 at 11:09 +0100, Tim Deegan wrote:
> > > At 14:31 +0100 on 07 Aug (1375885891), Ian Campbell wrote:
> > > > --- a/xen/include/asm-arm/page.h
> > > > +++ b/xen/include/asm-arm/page.h
> > > > @@ -337,7 +337,15 @@ static inline int gva_to_ipa(vaddr_t va, paddr_t *paddr)
> > > >  #define second_table_offset(va) TABLE_OFFSET(second_linear_offset(va))
> > > >  #define third_table_offset(va)  TABLE_OFFSET(third_linear_offset(va))
> > > >  
> > > > -#define clear_page(page)memset((void *)(page), 0, PAGE_SIZE)
> > > > +#ifdef CONFIG_ARM_64
> > > > +#define ZEROETH_SHIFT  (FIRST_SHIFT + LPAE_SHIFT)
> > > > +#define ZEROETH_SIZE   (1u << ZEROETH_SHIFT)
> > > > +#define ZEROETH_MASK   (~(ZEROETH_SIZE - 1))
> > > > +#define zeroeth_linear_offset(va) ((va) >> ZEROETH_SHIFT)
> > > > +#define zeroeth_table_offset(va)  TABLE_OFFSET(zeroeth_linear_offset(va))
> > > > +#endif
> > > 
> > > I'd prefer to see these defined alongside the other levels.  There's no
> > > harm in defining them on 32-bit builds.
> > 
> > My worry was that they would get used on 32-bit code paths by mistake,
> > which is a build error with this version but a (possibly obscure)
> > runtime error with the alternative.
> 
> Isn't at a build error to shift 32-bit values by 39 bits?

Good point, without checking I htink it is. Lets go with your suggested
change then.

>   I guess is
> more reliable, though; stet.
> 
> Tim.
> 
> > > Also, please update the comment above LPAE_SHIFT to describe the 4-level alternative. 
> > 
> > Will do.
> > 
> > 
> > 
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xen.org
> > http://lists.xen.org/xen-devel
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

[PATCH v2 03/10] xen: arm: Rename page table "hint" field to slightly more descriptive "contig"

[Ian Campbell]

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
 xen/arch/arm/mm.c          |    2 +-
 xen/include/asm-arm/page.h |    4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c
index f301e65..a4d65d8 100644
--- a/xen/arch/arm/mm.c
+++ b/xen/arch/arm/mm.c
@@ -526,7 +526,7 @@ static void __init create_mappings(unsigned long virt,
     count = nr_mfns / LPAE_ENTRIES;
     p = xen_second + second_linear_offset(virt);
     pte = mfn_to_xen_entry(base_mfn);
-    pte.pt.hint = 1;  /* These maps are in 16-entry contiguous chunks. */
+    pte.pt.contig = 1;  /* These maps are in 16-entry contiguous chunks. */
     for ( i = 0; i < count; i++ )
     {
         write_pte(p + i, pte);
diff --git a/xen/include/asm-arm/page.h b/xen/include/asm-arm/page.h
index 6317090..e34a8d6 100644
--- a/xen/include/asm-arm/page.h
+++ b/xen/include/asm-arm/page.h
@@ -115,7 +115,7 @@ typedef struct {
 
     /* These seven bits are only used in Block entries and are ignored
      * in Table entries. */
-    unsigned long hint:1;       /* In a block of 16 contiguous entries */
+    unsigned long contig:1;     /* In a block of 16 contiguous entries */
     unsigned long pxn:1;        /* Privileged-XN */
     unsigned long xn:1;         /* eXecute-Never */
     unsigned long avail:4;      /* Ignored by hardware */
@@ -150,7 +150,7 @@ typedef struct {
 
     /* These seven bits are only used in Block entries and are ignored
      * in Table entries. */
-    unsigned long hint:1;       /* In a block of 16 contiguous entries */
+    unsigned long contig:1;     /* In a block of 16 contiguous entries */
     unsigned long sbz2:1;
     unsigned long xn:1;         /* eXecute-Never */
     unsigned long avail:4;      /* Ignored by hardware */
-- 
1.7.2.5

[PATCH v2 04/10] xen: arm: refactor create_mappings

[Ian Campbell]

Allow it to work on contiguous second level tables mapping an arbitrary region
of memory. Rename it to create_32mb_mappings.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
 xen/arch/arm/mm.c |   19 +++++++++++--------
 1 files changed, 11 insertions(+), 8 deletions(-)

diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c
index a4d65d8..1ad8fe3 100644
--- a/xen/arch/arm/mm.c
+++ b/xen/arch/arm/mm.c
@@ -511,20 +511,23 @@ void __cpuinit mmu_init_secondary_cpu(void)
 }
 
 /* Create Xen's mappings of memory.
- * Base and virt must be 32MB aligned and size a multiple of 32MB. */
-static void __init create_mappings(unsigned long virt,
-                                   unsigned long base_mfn,
-                                   unsigned long nr_mfns)
+ * Base and virt must be 32MB aligned and size a multiple of 32MB.
+ * second must be a contiguous set of second level page tables
+ * covering the region starting at virt_offset. */
+static void __init create_32mb_mappings(lpae_t *second,
+                                        unsigned long virt_offset,
+                                        unsigned long base_mfn,
+                                        unsigned long nr_mfns)
 {
     unsigned long i, count;
     lpae_t pte, *p;
 
-    ASSERT(!((virt >> PAGE_SHIFT) % (16 * LPAE_ENTRIES)));
+    ASSERT(!((virt_offset >> PAGE_SHIFT) % (16 * LPAE_ENTRIES)));
     ASSERT(!(base_mfn % (16 * LPAE_ENTRIES)));
     ASSERT(!(nr_mfns % (16 * LPAE_ENTRIES)));
 
     count = nr_mfns / LPAE_ENTRIES;
-    p = xen_second + second_linear_offset(virt);
+    p = second + second_linear_offset(virt_offset);
     pte = mfn_to_xen_entry(base_mfn);
     pte.pt.contig = 1;  /* These maps are in 16-entry contiguous chunks. */
     for ( i = 0; i < count; i++ )
@@ -539,7 +542,7 @@ static void __init create_mappings(unsigned long virt,
 void __init setup_xenheap_mappings(unsigned long base_mfn,
                                    unsigned long nr_mfns)
 {
-    create_mappings(XENHEAP_VIRT_START, base_mfn, nr_mfns);
+    create_32mb_mappings(xen_second, XENHEAP_VIRT_START, base_mfn, nr_mfns);
 
     /* Record where the xenheap is, for translation routines. */
     xenheap_virt_end = XENHEAP_VIRT_START + nr_mfns * PAGE_SIZE;
@@ -559,7 +562,7 @@ void __init setup_frametable_mappings(paddr_t ps, paddr_t pe)
     /* Round up to 32M boundary */
     frametable_size = (frametable_size + 0x1ffffff) & ~0x1ffffff;
     base_mfn = alloc_boot_pages(frametable_size >> PAGE_SHIFT, 32<<(20-12));
-    create_mappings(FRAMETABLE_VIRT_START, base_mfn, frametable_size >> PAGE_SHIFT);
+    create_32mb_mappings(xen_second, FRAMETABLE_VIRT_START, base_mfn, frametable_size >> PAGE_SHIFT);
 
     memset(&frame_table[0], 0, nr_pages * sizeof(struct page_info));
     memset(&frame_table[nr_pages], -1,
-- 
1.7.2.5

[PATCH v2 05/10] xen: arm: define a macro to get this CPUs page table root

[Ian Campbell]

In a future patch the 32- and 64-bit root page tables will differ.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
---
 xen/arch/arm/mm.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c
index 1ad8fe3..7781202 100644
--- a/xen/arch/arm/mm.c
+++ b/xen/arch/arm/mm.c
@@ -60,6 +60,7 @@ lpae_t boot_first[LPAE_ENTRIES] __attribute__((__aligned__(4096)));
 /* Per-CPU pagetable pages */
 /* xen_pgtable == root of the trie (zeroeth level on 64-bit, first on 32-bit) */
 static DEFINE_PER_CPU(lpae_t *, xen_pgtable);
+#define THIS_CPU_PGTABLE this_cpu(xen_pgtable)
 /* xen_dommap == pages used by map_domain_page, these pages contain
  * the second level pagetables which mapp the domheap region
  * DOMHEAP_VIRT_START...DOMHEAP_VIRT_END in 2MB chunks. */
@@ -147,7 +148,7 @@ done:
 void dump_hyp_walk(vaddr_t addr)
 {
     uint64_t ttbr = READ_SYSREG64(TTBR0_EL2);
-    lpae_t *pgtable = this_cpu(xen_pgtable);
+    lpae_t *pgtable = THIS_CPU_PGTABLE;
 
     printk("Walking Hypervisor VA 0x%"PRIvaddr" "
            "on CPU%d via TTBR 0x%016"PRIx64"\n",
@@ -502,7 +503,7 @@ void __cpuinit mmu_init_secondary_cpu(void)
     uint64_t ttbr;
 
     /* Change to this CPU's pagetables */
-    ttbr = (uintptr_t)virt_to_maddr(this_cpu(xen_pgtable));
+    ttbr = (uintptr_t)virt_to_maddr(THIS_CPU_PGTABLE);
     WRITE_TTBR(ttbr);
 
     /* From now on, no mapping may be both writable and executable. */
-- 
1.7.2.5

Re: [PATCH v2 05/10] xen: arm: define a macro to get this CPUs page table root

[Stefano Stabellini]

On Wed, 7 Aug 2013, Ian Campbell wrote:
> In a future patch the 32- and 64-bit root page tables will differ.
> 
> Signed-off-by: Ian Campbell <ian.campbell@citrix.com>

Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>


>  xen/arch/arm/mm.c |    5 +++--
>  1 files changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c
> index 1ad8fe3..7781202 100644
> --- a/xen/arch/arm/mm.c
> +++ b/xen/arch/arm/mm.c
> @@ -60,6 +60,7 @@ lpae_t boot_first[LPAE_ENTRIES] __attribute__((__aligned__(4096)));
>  /* Per-CPU pagetable pages */
>  /* xen_pgtable == root of the trie (zeroeth level on 64-bit, first on 32-bit) */
>  static DEFINE_PER_CPU(lpae_t *, xen_pgtable);
> +#define THIS_CPU_PGTABLE this_cpu(xen_pgtable)
>  /* xen_dommap == pages used by map_domain_page, these pages contain
>   * the second level pagetables which mapp the domheap region
>   * DOMHEAP_VIRT_START...DOMHEAP_VIRT_END in 2MB chunks. */
> @@ -147,7 +148,7 @@ done:
>  void dump_hyp_walk(vaddr_t addr)
>  {
>      uint64_t ttbr = READ_SYSREG64(TTBR0_EL2);
> -    lpae_t *pgtable = this_cpu(xen_pgtable);
> +    lpae_t *pgtable = THIS_CPU_PGTABLE;
>  
>      printk("Walking Hypervisor VA 0x%"PRIvaddr" "
>             "on CPU%d via TTBR 0x%016"PRIx64"\n",
> @@ -502,7 +503,7 @@ void __cpuinit mmu_init_secondary_cpu(void)
>      uint64_t ttbr;
>  
>      /* Change to this CPU's pagetables */
> -    ttbr = (uintptr_t)virt_to_maddr(this_cpu(xen_pgtable));
> +    ttbr = (uintptr_t)virt_to_maddr(THIS_CPU_PGTABLE);
>      WRITE_TTBR(ttbr);
>  
>      /* From now on, no mapping may be both writable and executable. */
> -- 
> 1.7.2.5
> 

[PATCH v2 06/10] xen: arm: Add a pte_of_xenaddr helper

[Ian Campbell]

The open coded version is pretty ugly, not helped by rigid enforcement of an
80 character line length.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
 xen/arch/arm/mm.c |   13 +++++++++----
 1 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c
index 7781202..974d6df 100644
--- a/xen/arch/arm/mm.c
+++ b/xen/arch/arm/mm.c
@@ -331,6 +331,13 @@ void __cpuinit setup_virt_paging(void)
      * write to TTBR0 has completed. */                                 \
     flush_xen_text_tlb()
 
+static inline lpae_t pte_of_xenaddr(vaddr_t va)
+{
+    paddr_t ma = va + phys_offset;
+    unsigned long mfn = ma >> PAGE_SHIFT;
+    return mfn_to_xen_entry(mfn);
+}
+
 /* Boot-time pagetable setup.
  * Changes here may need matching changes in head.S */
 void __init setup_pagetables(unsigned long boot_phys_offset, paddr_t xen_paddr)
@@ -387,8 +394,7 @@ void __init setup_pagetables(unsigned long boot_phys_offset, paddr_t xen_paddr)
     flush_xen_text_tlb();
 
     /* Link in the fixmap pagetable */
-    pte = mfn_to_xen_entry((((unsigned long) xen_fixmap) + phys_offset)
-                           >> PAGE_SHIFT);
+    pte = pte_of_xenaddr((vaddr_t)xen_fixmap);
     pte.pt.table = 1;
     write_pte(xen_second + second_table_offset(FIXMAP_ADDR(0)), pte);
     /*
@@ -415,8 +421,7 @@ void __init setup_pagetables(unsigned long boot_phys_offset, paddr_t xen_paddr)
         write_pte(xen_xenmap + i, pte);
         /* No flush required here as page table is not hooked in yet. */
     }
-    pte = mfn_to_xen_entry((((unsigned long) xen_xenmap) + phys_offset)
-                           >> PAGE_SHIFT);
+    pte = pte_of_xenaddr((vaddr_t)xen_xenmap);
     pte.pt.table = 1;
     write_pte(xen_second + second_linear_offset(XEN_VIRT_START), pte);
     /* TLBFLUSH and ISB would be needed here, but wait until we set WXN */
-- 
1.7.2.5

[PATCH v2 07/10] xen: arm: fix is_xen_fixed_mfn

[Ian Campbell]

The current implementation is nonsense since the xenheap and the xen
text/data/etc mappings are nowhere near each other.

This is only actually used by the page offlining code, which isn't active on
ARM.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
 xen/include/asm-arm/mm.h |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/xen/include/asm-arm/mm.h b/xen/include/asm-arm/mm.h
index 5e7c5a3..baaf9c3 100644
--- a/xen/include/asm-arm/mm.h
+++ b/xen/include/asm-arm/mm.h
@@ -123,7 +123,9 @@ extern unsigned long xenheap_virt_end;
     unsigned long _mfn = (mfn);                                 \
     (_mfn >= xenheap_mfn_start && _mfn < xenheap_mfn_end);      \
 })
-#define is_xen_fixed_mfn(mfn) is_xen_heap_mfn(mfn)
+#define is_xen_fixed_mfn(mfn)                                   \
+    ((((mfn) << PAGE_SHIFT) >= virt_to_maddr(&_start)) &&       \
+     (((mfn) << PAGE_SHIFT) <= virt_to_maddr(&_end)))
 
 #define page_get_owner(_p)    (_p)->v.inuse.domain
 #define page_set_owner(_p,_d) ((_p)->v.inuse.domain = (_d))
-- 
1.7.2.5

[PATCH v2 08/10] xen: arm: allow virt_to_maddr to take either a pointer or an integer

[Ian Campbell]

This seems to be expected by common code which passes both pointers and
unsigned long as virtual addresses. The latter case in particular is in
init_node_heap() under a DIRECTMAP_VIRT_END #ifdef, which is why it hasn't
affected us yet (but will in a subsequent patch).

The new prototypes match the x86 versions apart from using vaddr_t instead of
unsigned long.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
 xen/include/asm-arm/mm.h |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/xen/include/asm-arm/mm.h b/xen/include/asm-arm/mm.h
index baaf9c3..7aca836 100644
--- a/xen/include/asm-arm/mm.h
+++ b/xen/include/asm-arm/mm.h
@@ -208,11 +208,12 @@ static inline void __iomem *ioremap_wc(paddr_t start, size_t len)
 #define paddr_to_pdx(pa)    pfn_to_pdx(paddr_to_pfn(pa))
 
 
-static inline paddr_t virt_to_maddr(const void *va)
+static inline paddr_t __virt_to_maddr(vaddr_t va)
 {
-    uint64_t par = va_to_par((vaddr_t)va);
-    return (par & PADDR_MASK & PAGE_MASK) | ((vaddr_t) va & ~PAGE_MASK);
+    uint64_t par = va_to_par(va);
+    return (par & PADDR_MASK & PAGE_MASK) | (va & ~PAGE_MASK);
 }
+#define virt_to_maddr(va)   __virt_to_maddr((vaddr_t)(va))
 
 static inline void *maddr_to_virt(paddr_t ma)
 {
-- 
1.7.2.5

[PATCH v2 09/10] xen: gate split heap code on its own config option rather than !X86

[Ian Campbell]

I'm going to want to disable this for 64 bit ARM.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: keir@xen.org
---
 xen/common/page_alloc.c      |    2 +-
 xen/include/asm-arm/config.h |    1 +
 2 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
index 25a7d3d..41251b2 100644
--- a/xen/common/page_alloc.c
+++ b/xen/common/page_alloc.c
@@ -1305,7 +1305,7 @@ void __init scrub_heap_pages(void)
  * XEN-HEAP SUB-ALLOCATOR
  */
 
-#if !defined(CONFIG_X86)
+#if defined(CONFIG_SEPARATE_XENHEAP)
 
 void init_xenheap_pages(paddr_t ps, paddr_t pe)
 {
diff --git a/xen/include/asm-arm/config.h b/xen/include/asm-arm/config.h
index 7e5daa0..fb9e93c 100644
--- a/xen/include/asm-arm/config.h
+++ b/xen/include/asm-arm/config.h
@@ -36,6 +36,7 @@
 #define CONFIG_SMP 1
 
 #define CONFIG_DOMAIN_PAGE 1
+#define CONFIG_SEPARATE_XENHEAP 1
 
 #define CONFIG_VIDEO 1
 
-- 
1.7.2.5

Re: [PATCH v2 09/10] xen: gate split heap code on its own config option rather than !X86

[Keir Fraser]

On 07/08/2013 14:31, "Ian Campbell" <ian.campbell@citrix.com> wrote:

> I'm going to want to disable this for 64 bit ARM.
> 
> Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
> Acked-by: Jan Beulich <jbeulich@suse.com>
> Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
> Cc: keir@xen.org

Acked-by: Keir Fraser <keir@xen.org>

> ---
>  xen/common/page_alloc.c      |    2 +-
>  xen/include/asm-arm/config.h |    1 +
>  2 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
> index 25a7d3d..41251b2 100644
> --- a/xen/common/page_alloc.c
> +++ b/xen/common/page_alloc.c
> @@ -1305,7 +1305,7 @@ void __init scrub_heap_pages(void)
>   * XEN-HEAP SUB-ALLOCATOR
>   */
>  
> -#if !defined(CONFIG_X86)
> +#if defined(CONFIG_SEPARATE_XENHEAP)
>  
>  void init_xenheap_pages(paddr_t ps, paddr_t pe)
>  {
> diff --git a/xen/include/asm-arm/config.h b/xen/include/asm-arm/config.h
> index 7e5daa0..fb9e93c 100644
> --- a/xen/include/asm-arm/config.h
> +++ b/xen/include/asm-arm/config.h
> @@ -36,6 +36,7 @@
>  #define CONFIG_SMP 1
>  
>  #define CONFIG_DOMAIN_PAGE 1
> +#define CONFIG_SEPARATE_XENHEAP 1
>  
>  #define CONFIG_VIDEO 1
>  

[PATCH v2 10/10] xen: arm: Use a direct mapping of RAM on arm64

[Ian Campbell]

We have plenty of virtual address space so we can avoid needing to map and
unmap pages all the time.

A totally arbitrarily chosen 32GB frame table leads to support for 5TB of RAM.
I haven't tested with anything near that amount of RAM though. There is plenty
of room to expand further when that becomes necessary.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
---
v2: Remove obsolete comments hung over from the 32-bit version
    Reformat TODO item about pte.pt.contig
    Duplicate prototype for alternate functions completely inside appropriate
    ifdef instead of just ifdeffing the body.
---
 xen/arch/arm/mm.c            |  157 ++++++++++++++++++++++++++++++++++--------
 xen/arch/arm/setup.c         |   77 ++++++++++++++++++++
 xen/include/asm-arm/config.h |   96 +++++++++++++++++++-------
 xen/include/asm-arm/mm.h     |   16 ++++
 4 files changed, 293 insertions(+), 53 deletions(-)

diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c
index 974d6df..69c157a 100644
--- a/xen/arch/arm/mm.c
+++ b/xen/arch/arm/mm.c
@@ -37,6 +37,7 @@
 #include <public/memory.h>
 #include <xen/sched.h>
 #include <xen/vmap.h>
+#include <asm/early_printk.h>
 #include <xsm/xsm.h>
 #include <xen/pfn.h>
 
@@ -48,6 +49,14 @@ struct domain *dom_xen, *dom_io, *dom_cow;
 lpae_t boot_pgtable[LPAE_ENTRIES] __attribute__((__aligned__(4096)));
 #ifdef CONFIG_ARM_64
 lpae_t boot_first[LPAE_ENTRIES] __attribute__((__aligned__(4096)));
+/* The first page of the first level mapping of the xenheap. The
+ * subsequent xenheap first level pages are dynamically allocated, but
+ * we need this one to bootstrap ourselves. */
+lpae_t xenheap_first_first[LPAE_ENTRIES] __attribute__((__aligned__(4096)));
+/* The zeroeth level slot which uses xenheap_first_first. Used because
+ * setup_xenheap_mappings otherwise relies on mfn_to_virt which isn't
+ * valid for a non-xenheap mapping. */
+static __initdata int xenheap_first_first_slot = -1;
 #endif
 
 /*
@@ -57,6 +66,9 @@ lpae_t boot_first[LPAE_ENTRIES] __attribute__((__aligned__(4096)));
  * xen_second, xen_fixmap and xen_xenmap are shared between all PCPUs.
  */
 
+#ifdef CONFIG_ARM_64
+#define THIS_CPU_PGTABLE boot_pgtable
+#else
 /* Per-CPU pagetable pages */
 /* xen_pgtable == root of the trie (zeroeth level on 64-bit, first on 32-bit) */
 static DEFINE_PER_CPU(lpae_t *, xen_pgtable);
@@ -65,6 +77,7 @@ static DEFINE_PER_CPU(lpae_t *, xen_pgtable);
  * the second level pagetables which mapp the domheap region
  * DOMHEAP_VIRT_START...DOMHEAP_VIRT_END in 2MB chunks. */
 static DEFINE_PER_CPU(lpae_t *, xen_dommap);
+#endif
 
 /* Common pagetable leaves */
 /* Second level page tables.
@@ -73,10 +86,16 @@ static DEFINE_PER_CPU(lpae_t *, xen_dommap);
  * addresses from 0 to 0x7fffffff. Offsets into it are calculated
  * with second_linear_offset(), not second_table_offset().
  *
- * Addresses 0x80000000 to 0xffffffff are covered by the per-cpu
- * xen_domheap mappings described above. However we allocate 4 pages
+ * On 32bit addresses 0x80000000 to 0xffffffff are covered by the
+ * per-cpu xen_domheap mappings described above. We allocate 4 pages
  * here for use in the boot page tables and the second two pages
  * become the boot CPUs xen_dommap pages.
+ *
+ * On 64bit addresses 0x80000000 to 0xffffffff are unused. However we
+ * allocate 4 pages here for use while relocating Xen, which currently
+ * expects a second level page to exist for all addresses in the first
+ * 4GB. We need to keep these extra mappings in place for seconary CPU
+ * bring up too. For now we just leave them forever.
  */
 lpae_t xen_second[LPAE_ENTRIES*4] __attribute__((__aligned__(4096*4)));
 /* First level page table used for fixmap */
@@ -92,7 +111,7 @@ uint64_t boot_ttbr;
 static paddr_t phys_offset;
 
 /* Limits of the Xen heap */
-unsigned long xenheap_mfn_start __read_mostly;
+unsigned long xenheap_mfn_start __read_mostly = ~0UL;
 unsigned long xenheap_mfn_end __read_mostly;
 unsigned long xenheap_virt_end __read_mostly;
 
@@ -112,7 +131,9 @@ static inline void check_memory_layout_alignment_constraints(void) {
     BUILD_BUG_ON(BOOT_MISC_VIRT_START & ~SECOND_MASK);
     /* 1GB aligned regions */
     BUILD_BUG_ON(XENHEAP_VIRT_START & ~FIRST_MASK);
+#ifdef CONFIG_DOMAIN_PAGE
     BUILD_BUG_ON(DOMHEAP_VIRT_START & ~FIRST_MASK);
+#endif
 }
 
 void dump_pt_walk(lpae_t *first, paddr_t addr)
@@ -180,6 +201,7 @@ void clear_fixmap(unsigned map)
     flush_xen_data_tlb_range_va(FIXMAP_ADDR(map), PAGE_SIZE);
 }
 
+#ifdef CONFIG_DOMAIN_PAGE
 void *map_domain_page_global(unsigned long mfn)
 {
     return vmap(&mfn, 1);
@@ -284,6 +306,7 @@ unsigned long domain_page_map_to_mfn(const void *va)
 
     return map[slot].pt.base + offset;
 }
+#endif
 
 void __init arch_init_memory(void)
 {
@@ -431,6 +454,7 @@ void __init setup_pagetables(unsigned long boot_phys_offset, paddr_t xen_paddr)
     /* Flush everything after setting WXN bit. */
     flush_xen_text_tlb();
 
+#ifdef CONFIG_ARM_32
     per_cpu(xen_pgtable, 0) = boot_pgtable;
     per_cpu(xen_dommap, 0) = xen_second +
         second_linear_offset(DOMHEAP_VIRT_START);
@@ -440,43 +464,33 @@ void __init setup_pagetables(unsigned long boot_phys_offset, paddr_t xen_paddr)
     memset(this_cpu(xen_dommap), 0, DOMHEAP_SECOND_PAGES*PAGE_SIZE);
     flush_xen_dcache_va_range(this_cpu(xen_dommap),
                               DOMHEAP_SECOND_PAGES*PAGE_SIZE);
+#endif
 }
-
+#ifdef CONFIG_ARM_64
+int init_secondary_pagetables(int cpu)
+{
+    /* All CPUs share a single page table on 64 bit */
+    return 0;
+}
+#else
 int init_secondary_pagetables(int cpu)
 {
-    lpae_t *root, *first, *domheap, pte;
+    lpae_t *first, *domheap, pte;
     int i;
 
-    root = alloc_xenheap_page();
-#ifdef CONFIG_ARM_64
-    first = alloc_xenheap_page();
-#else
-    first = root; /* root == first level on 32-bit 3-level trie */
-#endif
+    first = alloc_xenheap_page(); /* root == first level on 32-bit 3-level trie */
     domheap = alloc_xenheap_pages(get_order_from_pages(DOMHEAP_SECOND_PAGES), 0);
 
-    if ( root == NULL || domheap == NULL || first == NULL )
+    if ( domheap == NULL || first == NULL )
     {
         printk("Not enough free memory for secondary CPU%d pagetables\n", cpu);
         free_xenheap_pages(domheap, get_order_from_pages(DOMHEAP_SECOND_PAGES));
-#ifdef CONFIG_ARM_64
         free_xenheap_page(first);
-#endif
-        free_xenheap_page(root);
         return -ENOMEM;
     }
 
     /* Initialise root pagetable from root of boot tables */
-    memcpy(root, boot_pgtable, PAGE_SIZE);
-
-#ifdef CONFIG_ARM_64
-    /* Initialise first pagetable from first level of boot tables, and
-     * hook into the new root. */
-    memcpy(first, boot_first, PAGE_SIZE);
-    pte = mfn_to_xen_entry(virt_to_mfn(first));
-    pte.pt.table = 1;
-    write_pte(root, pte);
-#endif
+    memcpy(first, boot_pgtable, PAGE_SIZE);
 
     /* Ensure the domheap has no stray mappings */
     memset(domheap, 0, DOMHEAP_SECOND_PAGES*PAGE_SIZE);
@@ -490,17 +504,15 @@ int init_secondary_pagetables(int cpu)
         write_pte(&first[first_table_offset(DOMHEAP_VIRT_START+i*FIRST_SIZE)], pte);
     }
 
-    flush_xen_dcache_va_range(root, PAGE_SIZE);
-#ifdef CONFIG_ARM_64
     flush_xen_dcache_va_range(first, PAGE_SIZE);
-#endif
     flush_xen_dcache_va_range(domheap, DOMHEAP_SECOND_PAGES*PAGE_SIZE);
 
-    per_cpu(xen_pgtable, cpu) = root;
+    per_cpu(xen_pgtable, cpu) = first;
     per_cpu(xen_dommap, cpu) = domheap;
 
     return 0;
 }
+#endif
 
 /* MMU setup for secondary CPUS (which already have paging enabled) */
 void __cpuinit mmu_init_secondary_cpu(void)
@@ -544,6 +556,7 @@ static void __init create_32mb_mappings(lpae_t *second,
     flush_xen_data_tlb();
 }
 
+#ifdef CONFIG_ARM_32
 /* Set up the xenheap: up to 1GB of contiguous, always-mapped memory. */
 void __init setup_xenheap_mappings(unsigned long base_mfn,
                                    unsigned long nr_mfns)
@@ -555,6 +568,73 @@ void __init setup_xenheap_mappings(unsigned long base_mfn,
     xenheap_mfn_start = base_mfn;
     xenheap_mfn_end = base_mfn + nr_mfns;
 }
+#else /* CONFIG_ARM_64 */
+void __init setup_xenheap_mappings(unsigned long base_mfn,
+                                   unsigned long nr_mfns)
+{
+    lpae_t *first, pte;
+    unsigned long offset, end_mfn;
+    vaddr_t vaddr;
+
+    /* First call sets the xenheap physical offset. */
+    if ( xenheap_mfn_start == ~0UL )
+        xenheap_mfn_start = base_mfn;
+
+    if ( base_mfn < xenheap_mfn_start )
+        early_panic("cannot add xenheap mapping at %lx below heap start %lx\n",
+                    base_mfn, xenheap_mfn_start);
+
+    end_mfn = base_mfn + nr_mfns;
+
+    /* Align to previous 1GB boundary */
+    base_mfn &= ~FIRST_MASK;
+
+    offset = base_mfn - xenheap_mfn_start;
+    vaddr = DIRECTMAP_VIRT_START + offset*PAGE_SIZE;
+
+    while ( base_mfn < end_mfn )
+    {
+        int slot = zeroeth_table_offset(vaddr);
+        lpae_t *p = &boot_pgtable[slot];
+
+        if ( p->pt.valid )
+        {
+            /* mfn_to_virt is not valid on the 1st 1st mfn, since it
+             * is not within the xenheap. */
+            first = slot == xenheap_first_first_slot ?
+                xenheap_first_first : mfn_to_virt(p->pt.base);
+        }
+        else if ( xenheap_first_first_slot == -1)
+        {
+            /* Use xenheap_first_first to bootstrap the mappings */
+            first = xenheap_first_first;
+
+            pte = pte_of_xenaddr((vaddr_t)xenheap_first_first);
+            pte.pt.table = 1;
+            write_pte(p, pte);
+
+            xenheap_first_first_slot = slot;
+        }
+        else
+        {
+            unsigned long first_mfn = alloc_boot_pages(1, 1);
+            pte = mfn_to_xen_entry(first_mfn);
+            pte.pt.table = 1;
+            write_pte(p, pte);
+            first = mfn_to_virt(first_mfn);
+        }
+
+        pte = mfn_to_xen_entry(base_mfn);
+        /* TODO: Set pte.pt.contig when appropriate. */
+        write_pte(&first[first_table_offset(vaddr)], pte);
+
+        base_mfn += FIRST_SIZE>>PAGE_SHIFT;
+        vaddr += FIRST_SIZE;
+    }
+
+    flush_xen_data_tlb();
+}
+#endif
 
 /* Map a frame table to cover physical addresses ps through pe */
 void __init setup_frametable_mappings(paddr_t ps, paddr_t pe)
@@ -562,13 +642,32 @@ void __init setup_frametable_mappings(paddr_t ps, paddr_t pe)
     unsigned long nr_pages = (pe - ps) >> PAGE_SHIFT;
     unsigned long frametable_size = nr_pages * sizeof(struct page_info);
     unsigned long base_mfn;
+#ifdef CONFIG_ARM_64
+    lpae_t *second, pte;
+    unsigned long nr_second, second_base;
+    int i;
+#endif
 
     frametable_base_mfn = ps >> PAGE_SHIFT;
 
     /* Round up to 32M boundary */
     frametable_size = (frametable_size + 0x1ffffff) & ~0x1ffffff;
     base_mfn = alloc_boot_pages(frametable_size >> PAGE_SHIFT, 32<<(20-12));
+
+#ifdef CONFIG_ARM_64
+    nr_second = frametable_size >> SECOND_SHIFT;
+    second_base = alloc_boot_pages(nr_second, 1);
+    second = mfn_to_virt(second_base);
+    for ( i = 0; i < nr_second; i++ )
+    {
+        pte = mfn_to_xen_entry(second_base + i);
+        pte.pt.table = 1;
+        write_pte(&boot_first[first_table_offset(FRAMETABLE_VIRT_START)+i], pte);
+    }
+    create_32mb_mappings(second, 0, base_mfn, frametable_size >> PAGE_SHIFT);
+#else
     create_32mb_mappings(xen_second, FRAMETABLE_VIRT_START, base_mfn, frametable_size >> PAGE_SHIFT);
+#endif
 
     memset(&frame_table[0], 0, nr_pages * sizeof(struct page_info));
     memset(&frame_table[nr_pages], -1,
diff --git a/xen/arch/arm/setup.c b/xen/arch/arm/setup.c
index 1ec5e38..a25e483 100644
--- a/xen/arch/arm/setup.c
+++ b/xen/arch/arm/setup.c
@@ -278,6 +278,7 @@ static paddr_t __init get_xen_paddr(void)
     return paddr;
 }
 
+#ifdef CONFIG_ARM_32
 static void __init setup_mm(unsigned long dtb_paddr, size_t dtb_size)
 {
     paddr_t ram_start;
@@ -402,6 +403,82 @@ static void __init setup_mm(unsigned long dtb_paddr, size_t dtb_size)
 
     end_boot_allocator();
 }
+#else /* CONFIG_ARM_64 */
+static void __init setup_mm(unsigned long dtb_paddr, size_t dtb_size)
+{
+    paddr_t ram_start = ~0;
+    paddr_t ram_end = 0;
+    int bank;
+    unsigned long  xenheap_pages = 0;
+    unsigned long dtb_pages;
+
+    total_pages = 0;
+    for ( bank = 0 ; bank < early_info.mem.nr_banks; bank++ )
+    {
+        paddr_t bank_start = early_info.mem.bank[bank].start;
+        paddr_t bank_size  = early_info.mem.bank[bank].size;
+        paddr_t  bank_end = bank_start + bank_size;
+        unsigned long bank_pages = bank_size >> PAGE_SHIFT;
+        paddr_t s, e;
+
+        total_pages += bank_pages;
+
+        if ( bank_start < ram_start )
+            ram_start = bank_start;
+        if ( bank_end > ram_end )
+            ram_end = bank_end;
+
+        xenheap_pages += (bank_size >> PAGE_SHIFT);
+
+        /* XXX we assume that the ram regions are ordered */
+        s = bank_start;
+        while ( s < bank_end )
+        {
+            paddr_t n = bank_end;
+
+            e = next_module(s, &n);
+
+            if ( e == ~(paddr_t)0 )
+            {
+                e = n = bank_end;
+            }
+
+            setup_xenheap_mappings(s>>PAGE_SHIFT, (e-s)>>PAGE_SHIFT);
+
+            xenheap_mfn_end = e;
+
+            init_boot_pages(s, e);
+            s = n;
+        }
+    }
+
+    xenheap_virt_end = XENHEAP_VIRT_START + ram_end - ram_start;
+    xenheap_mfn_start = ram_start >> PAGE_SHIFT;
+    xenheap_mfn_end = ram_end >> PAGE_SHIFT;
+    xenheap_max_mfn(xenheap_mfn_end);
+
+    /*
+     * Need enough mapped pages for copying the DTB.
+     *
+     * TODO: The DTB (and other payloads) are assumed to be towards
+     * the start of RAM.
+     */
+    dtb_pages = (dtb_size + PAGE_SIZE-1) >> PAGE_SHIFT;
+
+    /*
+     * Copy the DTB.
+     *
+     * TODO: handle other payloads too.
+     */
+    device_tree_flattened = mfn_to_virt(alloc_boot_pages(dtb_pages, 1));
+    copy_from_paddr(device_tree_flattened, dtb_paddr, dtb_size, BUFFERABLE);
+
+    setup_frametable_mappings(ram_start, ram_end);
+    max_page = PFN_DOWN(ram_end);
+
+    end_boot_allocator();
+}
+#endif
 
 size_t __read_mostly cacheline_bytes;
 
diff --git a/xen/include/asm-arm/config.h b/xen/include/asm-arm/config.h
index fb9e93c..259d4c6 100644
--- a/xen/include/asm-arm/config.h
+++ b/xen/include/asm-arm/config.h
@@ -35,9 +35,6 @@
 
 #define CONFIG_SMP 1
 
-#define CONFIG_DOMAIN_PAGE 1
-#define CONFIG_SEPARATE_XENHEAP 1
-
 #define CONFIG_VIDEO 1
 
 #define OPT_CONSOLE_STR "dtuart"
@@ -76,38 +73,89 @@
 #include <xen/const.h>
 
 /*
- * Memory layout:
- *  0  -   2M   Unmapped
- *  2M -   4M   Xen text, data, bss
- *  4M -   6M   Fixmap: special-purpose 4K mapping slots
- *  6M  -  8M   Early boot misc (see below)
- *
- * 32M - 128M   Frametable: 24 bytes per page for 16GB of RAM
- * 256M -  1G   VMAP: ioremap and early_ioremap use this virtual address
- *                    space
- *
- *  1G -   2G   Xenheap: always-mapped memory
- *  2G -   4G   Domheap: on-demand-mapped
+ * Common ARM32 and ARM64 layout:
+ *   0  -   2M   Unmapped
+ *   2M -   4M   Xen text, data, bss
+ *   4M -   6M   Fixmap: special-purpose 4K mapping slots
+ *   6M -   8M   Early boot misc (see below)
  *
  * The early boot misc area is used:
  *   - in head.S for the DTB for device_tree_early_init().
  *   - in setup_pagetables() when relocating Xen.
+ *
+ * ARM32 layout:
+ *   0  -   8M   <COMMON>
+ *
+ *  32M - 128M   Frametable: 24 bytes per page for 16GB of RAM
+ * 256M -   1G   VMAP: ioremap and early_ioremap use this virtual address
+ *                    space
+ *
+ *   1G -   2G   Xenheap: always-mapped memory
+ *   2G -   4G   Domheap: on-demand-mapped
+ *
+ * ARM64 layout:
+ * 0x0000000000000000 - 0x0000007fffffffff (512GB, L0 slot [0])
+ *   0  -   8M   <COMMON>
+ *
+ *   1G -   2G   VMAP: ioremap and early_ioremap
+ *
+ *  32G -  64G   Frametable: 24 bytes per page for 5.3TB of RAM
+ *
+ * 0x0000008000000000 - 0x00007fffffffffff (127.5TB, L0 slots [1..255])
+ *  Unused
+ *
+ * 0x0000800000000000 - 0x000084ffffffffff (5TB, L0 slots [256..265])
+ *  1:1 mapping of RAM
+ *
+ * 0x0000850000000000 - 0x0000ffffffffffff (123TB, L0 slots [266..511])
+ *  Unused
  */
 
-#define XEN_VIRT_START         _AC(0x00200000,UL)
-#define FIXMAP_ADDR(n)        (_AC(0x00400000,UL) + (n) * PAGE_SIZE)
-#define BOOT_MISC_VIRT_START   _AC(0x00600000,UL)
-#define FRAMETABLE_VIRT_START  _AC(0x02000000,UL)
-#define VMAP_VIRT_START        _AC(0x10000000,UL)
-#define XENHEAP_VIRT_START     _AC(0x40000000,UL)
-#define DOMHEAP_VIRT_START     _AC(0x80000000,UL)
-#define DOMHEAP_VIRT_END       _AC(0xffffffff,UL)
+#define XEN_VIRT_START         _AT(vaddr_t,0x00200000)
+#define FIXMAP_ADDR(n)        (_AT(vaddr_t,0x00400000) + (n) * PAGE_SIZE)
+#define BOOT_MISC_VIRT_START   _AT(vaddr_t,0x00600000)
 
-#define VMAP_VIRT_END          XENHEAP_VIRT_START
 #define HYPERVISOR_VIRT_START  XEN_VIRT_START
 
+#ifdef CONFIG_ARM_32
+
+#define CONFIG_DOMAIN_PAGE 1
+#define CONFIG_SEPARATE_XENHEAP 1
+
+#define FRAMETABLE_VIRT_START  _AT(vaddr_t,0x02000000)
+#define VMAP_VIRT_START  _AT(vaddr_t,0x10000000)
+#define XENHEAP_VIRT_START     _AT(vaddr_t,0x40000000)
+#define XENHEAP_VIRT_END       _AT(vaddr_t,0x7fffffff)
+#define DOMHEAP_VIRT_START     _AT(vaddr_t,0x80000000)
+#define DOMHEAP_VIRT_END       _AT(vaddr_t,0xffffffff)
+
+#define VMAP_VIRT_END    XENHEAP_VIRT_START
+
 #define DOMHEAP_ENTRIES        1024  /* 1024 2MB mapping slots */
 
+#else /* ARM_64 */
+
+#define SLOT0_ENTRY_BITS  39
+#define SLOT0(slot) (_AT(vaddr_t,slot) << SLOT0_ENTRY_BITS)
+#define SLOT0_ENTRY_SIZE  SLOT0(1)
+#define GB(_gb)     (_AC(_gb, UL) << 30)
+
+#define VMAP_VIRT_START  GB(1)
+#define VMAP_VIRT_END    (VMAP_VIRT_START + GB(1) - 1)
+
+#define FRAMETABLE_VIRT_START  GB(32)
+#define FRAMETABLE_VIRT_END    (FRAMETABLE_VIRT_START + GB(32) - 1)
+
+#define DIRECTMAP_VIRT_START   SLOT0(256)
+#define DIRECTMAP_SIZE         (SLOT0_ENTRY_SIZE * (265-256))
+#define DIRECTMAP_VIRT_END     (DIRECTMAP_VIRT_START + DIRECTMAP_SIZE - 1)
+
+#define XENHEAP_VIRT_START     DIRECTMAP_VIRT_START
+
+#define HYPERVISOR_VIRT_END    DIRECTMAP_VIRT_END
+
+#endif
+
 /* Number of domheap pagetable pages required at the second level (2MB mappings) */
 #define DOMHEAP_SECOND_PAGES ((DOMHEAP_VIRT_END - DOMHEAP_VIRT_START + 1) >> FIRST_SHIFT)
 
diff --git a/xen/include/asm-arm/mm.h b/xen/include/asm-arm/mm.h
index 7aca836..27284d0 100644
--- a/xen/include/asm-arm/mm.h
+++ b/xen/include/asm-arm/mm.h
@@ -118,11 +118,18 @@ struct page_info
 extern unsigned long xenheap_mfn_start, xenheap_mfn_end;
 extern unsigned long xenheap_virt_end;
 
+#ifdef CONFIG_ARM_32
 #define is_xen_heap_page(page) is_xen_heap_mfn(page_to_mfn(page))
 #define is_xen_heap_mfn(mfn) ({                                 \
     unsigned long _mfn = (mfn);                                 \
     (_mfn >= xenheap_mfn_start && _mfn < xenheap_mfn_end);      \
 })
+#else
+#define is_xen_heap_page(page) ((page)->count_info & PGC_xen_heap)
+#define is_xen_heap_mfn(mfn) \
+    (mfn_valid(mfn) && is_xen_heap_page(__mfn_to_page(mfn)))
+#endif
+
 #define is_xen_fixed_mfn(mfn)                                   \
     ((((mfn) << PAGE_SHIFT) >= virt_to_maddr(&_start)) &&       \
      (((mfn) << PAGE_SHIFT) <= virt_to_maddr(&_end)))
@@ -215,12 +222,21 @@ static inline paddr_t __virt_to_maddr(vaddr_t va)
 }
 #define virt_to_maddr(va)   __virt_to_maddr((vaddr_t)(va))
 
+#ifdef CONFIG_ARM_32
 static inline void *maddr_to_virt(paddr_t ma)
 {
     ASSERT(is_xen_heap_mfn(ma >> PAGE_SHIFT));
     ma -= pfn_to_paddr(xenheap_mfn_start);
     return (void *)(unsigned long) ma + XENHEAP_VIRT_START;
 }
+#else
+static inline void *maddr_to_virt(paddr_t ma)
+{
+    ASSERT((ma >> PAGE_SHIFT) < (DIRECTMAP_SIZE >> PAGE_SHIFT));
+    ma -= pfn_to_paddr(xenheap_mfn_start);
+    return (void *)(unsigned long) ma + DIRECTMAP_VIRT_START;
+}
+#endif
 
 static inline int gvirt_to_maddr(vaddr_t va, paddr_t *pa)
 {
-- 
1.7.2.5

Re: [PATCH v2 10/10] xen: arm: Use a direct mapping of RAM on arm64

[Stefano Stabellini]

On Wed, 7 Aug 2013, Ian Campbell wrote:
> We have plenty of virtual address space so we can avoid needing to map and
> unmap pages all the time.
> 
> A totally arbitrarily chosen 32GB frame table leads to support for 5TB of RAM.
> I haven't tested with anything near that amount of RAM though. There is plenty
> of room to expand further when that becomes necessary.
> 
> Signed-off-by: Ian Campbell <ian.campbell@citrix.com>

Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>


> v2: Remove obsolete comments hung over from the 32-bit version
>     Reformat TODO item about pte.pt.contig
>     Duplicate prototype for alternate functions completely inside appropriate
>     ifdef instead of just ifdeffing the body.
> ---
>  xen/arch/arm/mm.c            |  157 ++++++++++++++++++++++++++++++++++--------
>  xen/arch/arm/setup.c         |   77 ++++++++++++++++++++
>  xen/include/asm-arm/config.h |   96 +++++++++++++++++++-------
>  xen/include/asm-arm/mm.h     |   16 ++++
>  4 files changed, 293 insertions(+), 53 deletions(-)
> 
> diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c
> index 974d6df..69c157a 100644
> --- a/xen/arch/arm/mm.c
> +++ b/xen/arch/arm/mm.c
> @@ -37,6 +37,7 @@
>  #include <public/memory.h>
>  #include <xen/sched.h>
>  #include <xen/vmap.h>
> +#include <asm/early_printk.h>
>  #include <xsm/xsm.h>
>  #include <xen/pfn.h>
>  
> @@ -48,6 +49,14 @@ struct domain *dom_xen, *dom_io, *dom_cow;
>  lpae_t boot_pgtable[LPAE_ENTRIES] __attribute__((__aligned__(4096)));
>  #ifdef CONFIG_ARM_64
>  lpae_t boot_first[LPAE_ENTRIES] __attribute__((__aligned__(4096)));
> +/* The first page of the first level mapping of the xenheap. The
> + * subsequent xenheap first level pages are dynamically allocated, but
> + * we need this one to bootstrap ourselves. */
> +lpae_t xenheap_first_first[LPAE_ENTRIES] __attribute__((__aligned__(4096)));
> +/* The zeroeth level slot which uses xenheap_first_first. Used because
> + * setup_xenheap_mappings otherwise relies on mfn_to_virt which isn't
> + * valid for a non-xenheap mapping. */
> +static __initdata int xenheap_first_first_slot = -1;
>  #endif
>  
>  /*
> @@ -57,6 +66,9 @@ lpae_t boot_first[LPAE_ENTRIES] __attribute__((__aligned__(4096)));
>   * xen_second, xen_fixmap and xen_xenmap are shared between all PCPUs.
>   */
>  
> +#ifdef CONFIG_ARM_64
> +#define THIS_CPU_PGTABLE boot_pgtable
> +#else
>  /* Per-CPU pagetable pages */
>  /* xen_pgtable == root of the trie (zeroeth level on 64-bit, first on 32-bit) */
>  static DEFINE_PER_CPU(lpae_t *, xen_pgtable);
> @@ -65,6 +77,7 @@ static DEFINE_PER_CPU(lpae_t *, xen_pgtable);
>   * the second level pagetables which mapp the domheap region
>   * DOMHEAP_VIRT_START...DOMHEAP_VIRT_END in 2MB chunks. */
>  static DEFINE_PER_CPU(lpae_t *, xen_dommap);
> +#endif
>  
>  /* Common pagetable leaves */
>  /* Second level page tables.
> @@ -73,10 +86,16 @@ static DEFINE_PER_CPU(lpae_t *, xen_dommap);
>   * addresses from 0 to 0x7fffffff. Offsets into it are calculated
>   * with second_linear_offset(), not second_table_offset().
>   *
> - * Addresses 0x80000000 to 0xffffffff are covered by the per-cpu
> - * xen_domheap mappings described above. However we allocate 4 pages
> + * On 32bit addresses 0x80000000 to 0xffffffff are covered by the
> + * per-cpu xen_domheap mappings described above. We allocate 4 pages
>   * here for use in the boot page tables and the second two pages
>   * become the boot CPUs xen_dommap pages.
> + *
> + * On 64bit addresses 0x80000000 to 0xffffffff are unused. However we
> + * allocate 4 pages here for use while relocating Xen, which currently
> + * expects a second level page to exist for all addresses in the first
> + * 4GB. We need to keep these extra mappings in place for seconary CPU
> + * bring up too. For now we just leave them forever.
>   */
>  lpae_t xen_second[LPAE_ENTRIES*4] __attribute__((__aligned__(4096*4)));
>  /* First level page table used for fixmap */
> @@ -92,7 +111,7 @@ uint64_t boot_ttbr;
>  static paddr_t phys_offset;
>  
>  /* Limits of the Xen heap */
> -unsigned long xenheap_mfn_start __read_mostly;
> +unsigned long xenheap_mfn_start __read_mostly = ~0UL;
>  unsigned long xenheap_mfn_end __read_mostly;
>  unsigned long xenheap_virt_end __read_mostly;
>  
> @@ -112,7 +131,9 @@ static inline void check_memory_layout_alignment_constraints(void) {
>      BUILD_BUG_ON(BOOT_MISC_VIRT_START & ~SECOND_MASK);
>      /* 1GB aligned regions */
>      BUILD_BUG_ON(XENHEAP_VIRT_START & ~FIRST_MASK);
> +#ifdef CONFIG_DOMAIN_PAGE
>      BUILD_BUG_ON(DOMHEAP_VIRT_START & ~FIRST_MASK);
> +#endif
>  }
>  
>  void dump_pt_walk(lpae_t *first, paddr_t addr)
> @@ -180,6 +201,7 @@ void clear_fixmap(unsigned map)
>      flush_xen_data_tlb_range_va(FIXMAP_ADDR(map), PAGE_SIZE);
>  }
>  
> +#ifdef CONFIG_DOMAIN_PAGE
>  void *map_domain_page_global(unsigned long mfn)
>  {
>      return vmap(&mfn, 1);
> @@ -284,6 +306,7 @@ unsigned long domain_page_map_to_mfn(const void *va)
>  
>      return map[slot].pt.base + offset;
>  }
> +#endif
>  
>  void __init arch_init_memory(void)
>  {
> @@ -431,6 +454,7 @@ void __init setup_pagetables(unsigned long boot_phys_offset, paddr_t xen_paddr)
>      /* Flush everything after setting WXN bit. */
>      flush_xen_text_tlb();
>  
> +#ifdef CONFIG_ARM_32
>      per_cpu(xen_pgtable, 0) = boot_pgtable;
>      per_cpu(xen_dommap, 0) = xen_second +
>          second_linear_offset(DOMHEAP_VIRT_START);
> @@ -440,43 +464,33 @@ void __init setup_pagetables(unsigned long boot_phys_offset, paddr_t xen_paddr)
>      memset(this_cpu(xen_dommap), 0, DOMHEAP_SECOND_PAGES*PAGE_SIZE);
>      flush_xen_dcache_va_range(this_cpu(xen_dommap),
>                                DOMHEAP_SECOND_PAGES*PAGE_SIZE);
> +#endif
>  }
> -
> +#ifdef CONFIG_ARM_64
> +int init_secondary_pagetables(int cpu)
> +{
> +    /* All CPUs share a single page table on 64 bit */
> +    return 0;
> +}
> +#else
>  int init_secondary_pagetables(int cpu)
>  {
> -    lpae_t *root, *first, *domheap, pte;
> +    lpae_t *first, *domheap, pte;
>      int i;
>  
> -    root = alloc_xenheap_page();
> -#ifdef CONFIG_ARM_64
> -    first = alloc_xenheap_page();
> -#else
> -    first = root; /* root == first level on 32-bit 3-level trie */
> -#endif
> +    first = alloc_xenheap_page(); /* root == first level on 32-bit 3-level trie */
>      domheap = alloc_xenheap_pages(get_order_from_pages(DOMHEAP_SECOND_PAGES), 0);
>  
> -    if ( root == NULL || domheap == NULL || first == NULL )
> +    if ( domheap == NULL || first == NULL )
>      {
>          printk("Not enough free memory for secondary CPU%d pagetables\n", cpu);
>          free_xenheap_pages(domheap, get_order_from_pages(DOMHEAP_SECOND_PAGES));
> -#ifdef CONFIG_ARM_64
>          free_xenheap_page(first);
> -#endif
> -        free_xenheap_page(root);
>          return -ENOMEM;
>      }
>  
>      /* Initialise root pagetable from root of boot tables */
> -    memcpy(root, boot_pgtable, PAGE_SIZE);
> -
> -#ifdef CONFIG_ARM_64
> -    /* Initialise first pagetable from first level of boot tables, and
> -     * hook into the new root. */
> -    memcpy(first, boot_first, PAGE_SIZE);
> -    pte = mfn_to_xen_entry(virt_to_mfn(first));
> -    pte.pt.table = 1;
> -    write_pte(root, pte);
> -#endif
> +    memcpy(first, boot_pgtable, PAGE_SIZE);
>  
>      /* Ensure the domheap has no stray mappings */
>      memset(domheap, 0, DOMHEAP_SECOND_PAGES*PAGE_SIZE);
> @@ -490,17 +504,15 @@ int init_secondary_pagetables(int cpu)
>          write_pte(&first[first_table_offset(DOMHEAP_VIRT_START+i*FIRST_SIZE)], pte);
>      }
>  
> -    flush_xen_dcache_va_range(root, PAGE_SIZE);
> -#ifdef CONFIG_ARM_64
>      flush_xen_dcache_va_range(first, PAGE_SIZE);
> -#endif
>      flush_xen_dcache_va_range(domheap, DOMHEAP_SECOND_PAGES*PAGE_SIZE);
>  
> -    per_cpu(xen_pgtable, cpu) = root;
> +    per_cpu(xen_pgtable, cpu) = first;
>      per_cpu(xen_dommap, cpu) = domheap;
>  
>      return 0;
>  }
> +#endif
>  
>  /* MMU setup for secondary CPUS (which already have paging enabled) */
>  void __cpuinit mmu_init_secondary_cpu(void)
> @@ -544,6 +556,7 @@ static void __init create_32mb_mappings(lpae_t *second,
>      flush_xen_data_tlb();
>  }
>  
> +#ifdef CONFIG_ARM_32
>  /* Set up the xenheap: up to 1GB of contiguous, always-mapped memory. */
>  void __init setup_xenheap_mappings(unsigned long base_mfn,
>                                     unsigned long nr_mfns)
> @@ -555,6 +568,73 @@ void __init setup_xenheap_mappings(unsigned long base_mfn,
>      xenheap_mfn_start = base_mfn;
>      xenheap_mfn_end = base_mfn + nr_mfns;
>  }
> +#else /* CONFIG_ARM_64 */
> +void __init setup_xenheap_mappings(unsigned long base_mfn,
> +                                   unsigned long nr_mfns)
> +{
> +    lpae_t *first, pte;
> +    unsigned long offset, end_mfn;
> +    vaddr_t vaddr;
> +
> +    /* First call sets the xenheap physical offset. */
> +    if ( xenheap_mfn_start == ~0UL )
> +        xenheap_mfn_start = base_mfn;
> +
> +    if ( base_mfn < xenheap_mfn_start )
> +        early_panic("cannot add xenheap mapping at %lx below heap start %lx\n",
> +                    base_mfn, xenheap_mfn_start);
> +
> +    end_mfn = base_mfn + nr_mfns;
> +
> +    /* Align to previous 1GB boundary */
> +    base_mfn &= ~FIRST_MASK;
> +
> +    offset = base_mfn - xenheap_mfn_start;
> +    vaddr = DIRECTMAP_VIRT_START + offset*PAGE_SIZE;
> +
> +    while ( base_mfn < end_mfn )
> +    {
> +        int slot = zeroeth_table_offset(vaddr);
> +        lpae_t *p = &boot_pgtable[slot];
> +
> +        if ( p->pt.valid )
> +        {
> +            /* mfn_to_virt is not valid on the 1st 1st mfn, since it
> +             * is not within the xenheap. */
> +            first = slot == xenheap_first_first_slot ?
> +                xenheap_first_first : mfn_to_virt(p->pt.base);
> +        }
> +        else if ( xenheap_first_first_slot == -1)
> +        {
> +            /* Use xenheap_first_first to bootstrap the mappings */
> +            first = xenheap_first_first;
> +
> +            pte = pte_of_xenaddr((vaddr_t)xenheap_first_first);
> +            pte.pt.table = 1;
> +            write_pte(p, pte);
> +
> +            xenheap_first_first_slot = slot;
> +        }
> +        else
> +        {
> +            unsigned long first_mfn = alloc_boot_pages(1, 1);
> +            pte = mfn_to_xen_entry(first_mfn);
> +            pte.pt.table = 1;
> +            write_pte(p, pte);
> +            first = mfn_to_virt(first_mfn);
> +        }
> +
> +        pte = mfn_to_xen_entry(base_mfn);
> +        /* TODO: Set pte.pt.contig when appropriate. */
> +        write_pte(&first[first_table_offset(vaddr)], pte);
> +
> +        base_mfn += FIRST_SIZE>>PAGE_SHIFT;
> +        vaddr += FIRST_SIZE;
> +    }
> +
> +    flush_xen_data_tlb();
> +}
> +#endif
>  
>  /* Map a frame table to cover physical addresses ps through pe */
>  void __init setup_frametable_mappings(paddr_t ps, paddr_t pe)
> @@ -562,13 +642,32 @@ void __init setup_frametable_mappings(paddr_t ps, paddr_t pe)
>      unsigned long nr_pages = (pe - ps) >> PAGE_SHIFT;
>      unsigned long frametable_size = nr_pages * sizeof(struct page_info);
>      unsigned long base_mfn;
> +#ifdef CONFIG_ARM_64
> +    lpae_t *second, pte;
> +    unsigned long nr_second, second_base;
> +    int i;
> +#endif
>  
>      frametable_base_mfn = ps >> PAGE_SHIFT;
>  
>      /* Round up to 32M boundary */
>      frametable_size = (frametable_size + 0x1ffffff) & ~0x1ffffff;
>      base_mfn = alloc_boot_pages(frametable_size >> PAGE_SHIFT, 32<<(20-12));
> +
> +#ifdef CONFIG_ARM_64
> +    nr_second = frametable_size >> SECOND_SHIFT;
> +    second_base = alloc_boot_pages(nr_second, 1);
> +    second = mfn_to_virt(second_base);
> +    for ( i = 0; i < nr_second; i++ )
> +    {
> +        pte = mfn_to_xen_entry(second_base + i);
> +        pte.pt.table = 1;
> +        write_pte(&boot_first[first_table_offset(FRAMETABLE_VIRT_START)+i], pte);
> +    }
> +    create_32mb_mappings(second, 0, base_mfn, frametable_size >> PAGE_SHIFT);
> +#else
>      create_32mb_mappings(xen_second, FRAMETABLE_VIRT_START, base_mfn, frametable_size >> PAGE_SHIFT);
> +#endif
>  
>      memset(&frame_table[0], 0, nr_pages * sizeof(struct page_info));
>      memset(&frame_table[nr_pages], -1,
> diff --git a/xen/arch/arm/setup.c b/xen/arch/arm/setup.c
> index 1ec5e38..a25e483 100644
> --- a/xen/arch/arm/setup.c
> +++ b/xen/arch/arm/setup.c
> @@ -278,6 +278,7 @@ static paddr_t __init get_xen_paddr(void)
>      return paddr;
>  }
>  
> +#ifdef CONFIG_ARM_32
>  static void __init setup_mm(unsigned long dtb_paddr, size_t dtb_size)
>  {
>      paddr_t ram_start;
> @@ -402,6 +403,82 @@ static void __init setup_mm(unsigned long dtb_paddr, size_t dtb_size)
>  
>      end_boot_allocator();
>  }
> +#else /* CONFIG_ARM_64 */
> +static void __init setup_mm(unsigned long dtb_paddr, size_t dtb_size)
> +{
> +    paddr_t ram_start = ~0;
> +    paddr_t ram_end = 0;
> +    int bank;
> +    unsigned long  xenheap_pages = 0;
> +    unsigned long dtb_pages;
> +
> +    total_pages = 0;
> +    for ( bank = 0 ; bank < early_info.mem.nr_banks; bank++ )
> +    {
> +        paddr_t bank_start = early_info.mem.bank[bank].start;
> +        paddr_t bank_size  = early_info.mem.bank[bank].size;
> +        paddr_t  bank_end = bank_start + bank_size;
> +        unsigned long bank_pages = bank_size >> PAGE_SHIFT;
> +        paddr_t s, e;
> +
> +        total_pages += bank_pages;
> +
> +        if ( bank_start < ram_start )
> +            ram_start = bank_start;
> +        if ( bank_end > ram_end )
> +            ram_end = bank_end;
> +
> +        xenheap_pages += (bank_size >> PAGE_SHIFT);
> +
> +        /* XXX we assume that the ram regions are ordered */
> +        s = bank_start;
> +        while ( s < bank_end )
> +        {
> +            paddr_t n = bank_end;
> +
> +            e = next_module(s, &n);
> +
> +            if ( e == ~(paddr_t)0 )
> +            {
> +                e = n = bank_end;
> +            }
> +
> +            setup_xenheap_mappings(s>>PAGE_SHIFT, (e-s)>>PAGE_SHIFT);
> +
> +            xenheap_mfn_end = e;
> +
> +            init_boot_pages(s, e);
> +            s = n;
> +        }
> +    }
> +
> +    xenheap_virt_end = XENHEAP_VIRT_START + ram_end - ram_start;
> +    xenheap_mfn_start = ram_start >> PAGE_SHIFT;
> +    xenheap_mfn_end = ram_end >> PAGE_SHIFT;
> +    xenheap_max_mfn(xenheap_mfn_end);
> +
> +    /*
> +     * Need enough mapped pages for copying the DTB.
> +     *
> +     * TODO: The DTB (and other payloads) are assumed to be towards
> +     * the start of RAM.
> +     */
> +    dtb_pages = (dtb_size + PAGE_SIZE-1) >> PAGE_SHIFT;
> +
> +    /*
> +     * Copy the DTB.
> +     *
> +     * TODO: handle other payloads too.
> +     */
> +    device_tree_flattened = mfn_to_virt(alloc_boot_pages(dtb_pages, 1));
> +    copy_from_paddr(device_tree_flattened, dtb_paddr, dtb_size, BUFFERABLE);
> +
> +    setup_frametable_mappings(ram_start, ram_end);
> +    max_page = PFN_DOWN(ram_end);
> +
> +    end_boot_allocator();
> +}
> +#endif
>  
>  size_t __read_mostly cacheline_bytes;
>  
> diff --git a/xen/include/asm-arm/config.h b/xen/include/asm-arm/config.h
> index fb9e93c..259d4c6 100644
> --- a/xen/include/asm-arm/config.h
> +++ b/xen/include/asm-arm/config.h
> @@ -35,9 +35,6 @@
>  
>  #define CONFIG_SMP 1
>  
> -#define CONFIG_DOMAIN_PAGE 1
> -#define CONFIG_SEPARATE_XENHEAP 1
> -
>  #define CONFIG_VIDEO 1
>  
>  #define OPT_CONSOLE_STR "dtuart"
> @@ -76,38 +73,89 @@
>  #include <xen/const.h>
>  
>  /*
> - * Memory layout:
> - *  0  -   2M   Unmapped
> - *  2M -   4M   Xen text, data, bss
> - *  4M -   6M   Fixmap: special-purpose 4K mapping slots
> - *  6M  -  8M   Early boot misc (see below)
> - *
> - * 32M - 128M   Frametable: 24 bytes per page for 16GB of RAM
> - * 256M -  1G   VMAP: ioremap and early_ioremap use this virtual address
> - *                    space
> - *
> - *  1G -   2G   Xenheap: always-mapped memory
> - *  2G -   4G   Domheap: on-demand-mapped
> + * Common ARM32 and ARM64 layout:
> + *   0  -   2M   Unmapped
> + *   2M -   4M   Xen text, data, bss
> + *   4M -   6M   Fixmap: special-purpose 4K mapping slots
> + *   6M -   8M   Early boot misc (see below)
>   *
>   * The early boot misc area is used:
>   *   - in head.S for the DTB for device_tree_early_init().
>   *   - in setup_pagetables() when relocating Xen.
> + *
> + * ARM32 layout:
> + *   0  -   8M   <COMMON>
> + *
> + *  32M - 128M   Frametable: 24 bytes per page for 16GB of RAM
> + * 256M -   1G   VMAP: ioremap and early_ioremap use this virtual address
> + *                    space
> + *
> + *   1G -   2G   Xenheap: always-mapped memory
> + *   2G -   4G   Domheap: on-demand-mapped
> + *
> + * ARM64 layout:
> + * 0x0000000000000000 - 0x0000007fffffffff (512GB, L0 slot [0])
> + *   0  -   8M   <COMMON>
> + *
> + *   1G -   2G   VMAP: ioremap and early_ioremap
> + *
> + *  32G -  64G   Frametable: 24 bytes per page for 5.3TB of RAM
> + *
> + * 0x0000008000000000 - 0x00007fffffffffff (127.5TB, L0 slots [1..255])
> + *  Unused
> + *
> + * 0x0000800000000000 - 0x000084ffffffffff (5TB, L0 slots [256..265])
> + *  1:1 mapping of RAM
> + *
> + * 0x0000850000000000 - 0x0000ffffffffffff (123TB, L0 slots [266..511])
> + *  Unused
>   */
>  
> -#define XEN_VIRT_START         _AC(0x00200000,UL)
> -#define FIXMAP_ADDR(n)        (_AC(0x00400000,UL) + (n) * PAGE_SIZE)
> -#define BOOT_MISC_VIRT_START   _AC(0x00600000,UL)
> -#define FRAMETABLE_VIRT_START  _AC(0x02000000,UL)
> -#define VMAP_VIRT_START        _AC(0x10000000,UL)
> -#define XENHEAP_VIRT_START     _AC(0x40000000,UL)
> -#define DOMHEAP_VIRT_START     _AC(0x80000000,UL)
> -#define DOMHEAP_VIRT_END       _AC(0xffffffff,UL)
> +#define XEN_VIRT_START         _AT(vaddr_t,0x00200000)
> +#define FIXMAP_ADDR(n)        (_AT(vaddr_t,0x00400000) + (n) * PAGE_SIZE)
> +#define BOOT_MISC_VIRT_START   _AT(vaddr_t,0x00600000)
>  
> -#define VMAP_VIRT_END          XENHEAP_VIRT_START
>  #define HYPERVISOR_VIRT_START  XEN_VIRT_START
>  
> +#ifdef CONFIG_ARM_32
> +
> +#define CONFIG_DOMAIN_PAGE 1
> +#define CONFIG_SEPARATE_XENHEAP 1
> +
> +#define FRAMETABLE_VIRT_START  _AT(vaddr_t,0x02000000)
> +#define VMAP_VIRT_START  _AT(vaddr_t,0x10000000)
> +#define XENHEAP_VIRT_START     _AT(vaddr_t,0x40000000)
> +#define XENHEAP_VIRT_END       _AT(vaddr_t,0x7fffffff)
> +#define DOMHEAP_VIRT_START     _AT(vaddr_t,0x80000000)
> +#define DOMHEAP_VIRT_END       _AT(vaddr_t,0xffffffff)
> +
> +#define VMAP_VIRT_END    XENHEAP_VIRT_START
> +
>  #define DOMHEAP_ENTRIES        1024  /* 1024 2MB mapping slots */
>  
> +#else /* ARM_64 */
> +
> +#define SLOT0_ENTRY_BITS  39
> +#define SLOT0(slot) (_AT(vaddr_t,slot) << SLOT0_ENTRY_BITS)
> +#define SLOT0_ENTRY_SIZE  SLOT0(1)
> +#define GB(_gb)     (_AC(_gb, UL) << 30)
> +
> +#define VMAP_VIRT_START  GB(1)
> +#define VMAP_VIRT_END    (VMAP_VIRT_START + GB(1) - 1)
> +
> +#define FRAMETABLE_VIRT_START  GB(32)
> +#define FRAMETABLE_VIRT_END    (FRAMETABLE_VIRT_START + GB(32) - 1)
> +
> +#define DIRECTMAP_VIRT_START   SLOT0(256)
> +#define DIRECTMAP_SIZE         (SLOT0_ENTRY_SIZE * (265-256))
> +#define DIRECTMAP_VIRT_END     (DIRECTMAP_VIRT_START + DIRECTMAP_SIZE - 1)
> +
> +#define XENHEAP_VIRT_START     DIRECTMAP_VIRT_START
> +
> +#define HYPERVISOR_VIRT_END    DIRECTMAP_VIRT_END
> +
> +#endif
> +
>  /* Number of domheap pagetable pages required at the second level (2MB mappings) */
>  #define DOMHEAP_SECOND_PAGES ((DOMHEAP_VIRT_END - DOMHEAP_VIRT_START + 1) >> FIRST_SHIFT)
>  
> diff --git a/xen/include/asm-arm/mm.h b/xen/include/asm-arm/mm.h
> index 7aca836..27284d0 100644
> --- a/xen/include/asm-arm/mm.h
> +++ b/xen/include/asm-arm/mm.h
> @@ -118,11 +118,18 @@ struct page_info
>  extern unsigned long xenheap_mfn_start, xenheap_mfn_end;
>  extern unsigned long xenheap_virt_end;
>  
> +#ifdef CONFIG_ARM_32
>  #define is_xen_heap_page(page) is_xen_heap_mfn(page_to_mfn(page))
>  #define is_xen_heap_mfn(mfn) ({                                 \
>      unsigned long _mfn = (mfn);                                 \
>      (_mfn >= xenheap_mfn_start && _mfn < xenheap_mfn_end);      \
>  })
> +#else
> +#define is_xen_heap_page(page) ((page)->count_info & PGC_xen_heap)
> +#define is_xen_heap_mfn(mfn) \
> +    (mfn_valid(mfn) && is_xen_heap_page(__mfn_to_page(mfn)))
> +#endif
> +
>  #define is_xen_fixed_mfn(mfn)                                   \
>      ((((mfn) << PAGE_SHIFT) >= virt_to_maddr(&_start)) &&       \
>       (((mfn) << PAGE_SHIFT) <= virt_to_maddr(&_end)))
> @@ -215,12 +222,21 @@ static inline paddr_t __virt_to_maddr(vaddr_t va)
>  }
>  #define virt_to_maddr(va)   __virt_to_maddr((vaddr_t)(va))
>  
> +#ifdef CONFIG_ARM_32
>  static inline void *maddr_to_virt(paddr_t ma)
>  {
>      ASSERT(is_xen_heap_mfn(ma >> PAGE_SHIFT));
>      ma -= pfn_to_paddr(xenheap_mfn_start);
>      return (void *)(unsigned long) ma + XENHEAP_VIRT_START;
>  }
> +#else
> +static inline void *maddr_to_virt(paddr_t ma)
> +{
> +    ASSERT((ma >> PAGE_SHIFT) < (DIRECTMAP_SIZE >> PAGE_SHIFT));
> +    ma -= pfn_to_paddr(xenheap_mfn_start);
> +    return (void *)(unsigned long) ma + DIRECTMAP_VIRT_START;
> +}
> +#endif
>  
>  static inline int gvirt_to_maddr(vaddr_t va, paddr_t *pa)
>  {
> -- 
> 1.7.2.5
>