From mboxrd@z Thu Jan  1 00:00:00 1970
From: "Jan Beulich" <JBeulich@suse.com>
Subject: [PATCH v2] x86: debugging code for testing 16Tb support
 on smaller memory systems
Date: Wed, 23 Jan 2013 14:26:28 +0000
Message-ID: <5100012402000078000B8AB9@nat28.tlf.novell.com>
References: <50FE7BF502000078000B82F8@nat28.tlf.novell.com>
	<50FE7EFD02000078000B8359@nat28.tlf.novell.com>
Mime-Version: 1.0
Content-Type: multipart/mixed; boundary="=__Part30019204.0__="
Return-path: <xen-devel-bounces@lists.xen.org>
In-Reply-To: <50FE7EFD02000078000B8359@nat28.tlf.novell.com>
List-Unsubscribe: <http://lists.xen.org/cgi-bin/mailman/options/xen-devel>,
	<mailto:xen-devel-request@lists.xen.org?subject=unsubscribe>
List-Post: <mailto:xen-devel@lists.xen.org>
List-Help: <mailto:xen-devel-request@lists.xen.org?subject=help>
List-Subscribe: <http://lists.xen.org/cgi-bin/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xen.org?subject=subscribe>
Sender: xen-devel-bounces@lists.xen.org
Errors-To: xen-devel-bounces@lists.xen.org
To: xen-devel <xen-devel@lists.xen.org>
List-Id: xen-devel@lists.xenproject.org

This is a MIME message. If you are reading this text, you may want to 
consider changing to a mail reader or gateway that understands how to 
properly handle MIME multipart messages.

--=__Part30019204.0__=
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

Signed-off-by: Jan Beulich <jbeulich@suse.com>
---
v2: Removed unwanted bits and switched to byte-granular "highmem-start"
    option.

--- a/docs/misc/xen-command-line.markdown
+++ b/docs/misc/xen-command-line.markdown
@@ -546,6 +546,12 @@ Paging (HAP).
 ### hvm\_port80
 > `=3D <boolean>`
=20
+### highmem-start
+> `=3D <size>`
+
+Specify the memory boundary past which memory will be treated as highmem =
(x86
+debug hypervisor only).
+
 ### idle\_latency\_factor
 > `=3D <integer>`
=20
--- a/xen/arch/x86/domain_page.c
+++ b/xen/arch/x86/domain_page.c
@@ -66,8 +66,10 @@ void *map_domain_page(unsigned long mfn)
     struct mapcache_vcpu *vcache;
     struct vcpu_maphash_entry *hashent;
=20
+#ifdef NDEBUG
     if ( mfn <=3D PFN_DOWN(__pa(HYPERVISOR_VIRT_END - 1)) )
         return mfn_to_virt(mfn);
+#endif
=20
     v =3D mapcache_current_vcpu();
     if ( !v || is_hvm_vcpu(v) )
@@ -249,8 +251,10 @@ int mapcache_domain_init(struct domain *
     if ( is_hvm_domain(d) || is_idle_domain(d) )
         return 0;
=20
+#ifdef NDEBUG
     if ( !mem_hotplug && max_page <=3D PFN_DOWN(__pa(HYPERVISOR_VIRT_END =
- 1)) )
         return 0;
+#endif
=20
     dcache->l1tab =3D xzalloc_array(l1_pgentry_t *, MAPCACHE_L2_ENTRIES + =
1);
     d->arch.perdomain_l2_pg[MAPCACHE_SLOT] =3D alloc_domheap_page(NULL, =
memf);
@@ -418,8 +422,10 @@ void *map_domain_page_global(unsigned lo
=20
     ASSERT(!in_irq() && local_irq_is_enabled());
=20
+#ifdef NDEBUG
     if ( mfn <=3D PFN_DOWN(__pa(HYPERVISOR_VIRT_END - 1)) )
         return mfn_to_virt(mfn);
+#endif
=20
     spin_lock(&globalmap_lock);
=20
--- a/xen/arch/x86/mm.c
+++ b/xen/arch/x86/mm.c
@@ -250,6 +250,14 @@ void __init init_frametable(void)
         init_spagetable();
 }
=20
+#ifndef NDEBUG
+static unsigned int __read_mostly root_pgt_pv_xen_slots
+    =3D ROOT_PAGETABLE_PV_XEN_SLOTS;
+static l4_pgentry_t __read_mostly split_l4e;
+#else
+#define root_pgt_pv_xen_slots ROOT_PAGETABLE_PV_XEN_SLOTS
+#endif
+
 void __init arch_init_memory(void)
 {
     unsigned long i, pfn, rstart_pfn, rend_pfn, iostart_pfn, ioend_pfn;
@@ -344,6 +352,40 @@ void __init arch_init_memory(void)
     efi_init_memory();
=20
     mem_sharing_init();
+
+#ifndef NDEBUG
+    if ( highmem_start )
+    {
+        unsigned long split_va =3D (unsigned long)__va(highmem_start);
+
+        if ( split_va < HYPERVISOR_VIRT_END &&
+             split_va - 1 =3D=3D (unsigned long)__va(highmem_start - 1) )
+        {
+            root_pgt_pv_xen_slots =3D l4_table_offset(split_va) -
+                                    ROOT_PAGETABLE_FIRST_XEN_SLOT;
+            ASSERT(root_pgt_pv_xen_slots < ROOT_PAGETABLE_PV_XEN_SLOTS);
+            if ( l4_table_offset(split_va) =3D=3D l4_table_offset(split_va=
 - 1) )
+            {
+                l3_pgentry_t *l3tab =3D alloc_xen_pagetable();
+
+                if ( l3tab )
+                {
+                    const l3_pgentry_t *l3idle =3D
+                        l4e_to_l3e(idle_pg_table[l4_table_offset(split_va)=
]);
+
+                    for ( i =3D 0; i < l3_table_offset(split_va); ++i )
+                        l3tab[i] =3D l3idle[i];
+                    for ( ; i <=3D L3_PAGETABLE_ENTRIES; ++i )
+                        l3tab[i] =3D l3e_empty();
+                    split_l4e =3D l4e_from_pfn(virt_to_mfn(l3tab),
+                                             __PAGE_HYPERVISOR);
+                }
+                else
+                    ++root_pgt_pv_xen_slots;
+            }
+        }
+    }
+#endif
 }
=20
 int page_is_ram_type(unsigned long mfn, unsigned long mem_type)
@@ -1320,7 +1362,12 @@ void init_guest_l4_table(l4_pgentry_t l4
     /* Xen private mappings. */
     memcpy(&l4tab[ROOT_PAGETABLE_FIRST_XEN_SLOT],
            &idle_pg_table[ROOT_PAGETABLE_FIRST_XEN_SLOT],
-           ROOT_PAGETABLE_PV_XEN_SLOTS * sizeof(l4_pgentry_t));
+           root_pgt_pv_xen_slots * sizeof(l4_pgentry_t));
+#ifndef NDEBUG
+    if ( l4e_get_intpte(split_l4e) )
+        l4tab[ROOT_PAGETABLE_FIRST_XEN_SLOT + root_pgt_pv_xen_slots] =3D
+            split_l4e;
+#endif
     l4tab[l4_table_offset(LINEAR_PT_VIRT_START)] =3D
         l4e_from_pfn(domain_page_map_to_mfn(l4tab), __PAGE_HYPERVISOR);
     l4tab[l4_table_offset(PERDOMAIN_VIRT_START)] =3D
--- a/xen/arch/x86/setup.c
+++ b/xen/arch/x86/setup.c
@@ -82,6 +82,11 @@ boolean_param("noapic", skip_ioapic_setu
 s8 __read_mostly xen_cpuidle =3D -1;
 boolean_param("cpuidle", xen_cpuidle);
=20
+#ifndef NDEBUG
+unsigned long __initdata highmem_start;
+size_param("highmem-start", highmem_start);
+#endif
+
 cpumask_t __read_mostly cpu_present_map;
=20
 unsigned long __read_mostly xen_phys_start;
@@ -787,6 +792,14 @@ void __init __start_xen(unsigned long mb
     modules_headroom =3D bzimage_headroom(bootstrap_map(mod), mod->mod_end=
);
     bootstrap_map(NULL);
=20
+#ifndef highmem_start
+    /* Don't allow split below 4Gb. */
+    if ( highmem_start < GB(4) )
+        highmem_start =3D 0;
+    else /* align to L3 entry boundary */
+        highmem_start &=3D ~((1UL << L3_PAGETABLE_SHIFT) - 1);
+#endif
+
     for ( i =3D boot_e820.nr_map-1; i >=3D 0; i-- )
     {
         uint64_t s, e, mask =3D (1UL << L2_PAGETABLE_SHIFT) - 1;
@@ -915,6 +928,9 @@ void __init __start_xen(unsigned long mb
             /* Don't overlap with other modules. */
             end =3D consider_modules(s, e, size, mod, mbi->mods_count, =
j);
=20
+            if ( highmem_start && end > highmem_start )
+                continue;
+
             if ( s < end &&
                  (headroom ||
                   ((end - size) >> PAGE_SHIFT) > mod[j].mod_start) )
@@ -956,6 +972,8 @@ void __init __start_xen(unsigned long mb
     kexec_reserve_area(&boot_e820);
=20
     setup_max_pdx();
+    if ( highmem_start )
+        xenheap_max_mfn(PFN_DOWN(highmem_start));
=20
     /*
      * Walk every RAM region and map it in its entirety (on x86/64, at =
least)
@@ -1127,7 +1145,8 @@ void __init __start_xen(unsigned long mb
         unsigned long limit =3D virt_to_mfn(HYPERVISOR_VIRT_END - 1);
         uint64_t mask =3D PAGE_SIZE - 1;
=20
-        xenheap_max_mfn(limit);
+        if ( !highmem_start )
+            xenheap_max_mfn(limit);
=20
         /* Pass the remaining memory to the allocator. */
         for ( i =3D 0; i < boot_e820.nr_map; i++ )
--- a/xen/common/page_alloc.c
+++ b/xen/common/page_alloc.c
@@ -45,6 +45,7 @@
 #include <asm/flushtlb.h>
 #ifdef CONFIG_X86
 #include <asm/p2m.h>
+#include <asm/setup.h> /* for highmem_start only */
 #else
 #define p2m_pod_offline_or_broken_hit(pg) 0
 #define p2m_pod_offline_or_broken_replace(pg) BUG_ON(pg !=3D NULL)
@@ -203,6 +204,25 @@ unsigned long __init alloc_boot_pages(
         pg =3D (r->e - nr_pfns) & ~(pfn_align - 1);
         if ( pg < r->s )
             continue;
+
+#if defined(CONFIG_X86) && !defined(NDEBUG)
+        /*
+         * Filtering pfn_align =3D=3D 1 since the only allocations using =
a bigger
+         * alignment are the ones used for setting up the frame table =
chunks.
+         * Those allocations get remapped anyway, i.e. them not having =
1:1
+         * mappings always accessible is not a problem.
+         */
+        if ( highmem_start && pfn_align =3D=3D 1 &&
+             r->e > PFN_DOWN(highmem_start) )
+        {
+            pg =3D r->s;
+            if ( pg + nr_pfns > PFN_DOWN(highmem_start) )
+                continue;
+            r->s =3D pg + nr_pfns;
+            return pg;
+        }
+#endif
+
         _e =3D r->e;
         r->e =3D pg;
         bootmem_region_add(pg + nr_pfns, _e);
--- a/xen/include/asm-x86/setup.h
+++ b/xen/include/asm-x86/setup.h
@@ -43,4 +43,10 @@ void microcode_grab_module(
=20
 extern uint8_t kbd_shift_flags;
=20
+#ifdef NDEBUG
+# define highmem_start 0
+#else
+extern unsigned long highmem_start;
+#endif
+
 #endif



--=__Part30019204.0__=
Content-Type: text/plain; name="x86-map-domain-debug.patch"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: attachment; filename="x86-map-domain-debug.patch"

x86: debugging code for testing 16Tb support on smaller memory systems=0A=
=0ASigned-off-by: Jan Beulich <jbeulich@suse.com>=0A---=0Av2: Removed =
unwanted bits and switched to byte-granular "highmem-start"=0A    =
option.=0A=0A--- a/docs/misc/xen-command-line.markdown=0A+++ b/docs/misc/xe=
n-command-line.markdown=0A@@ -546,6 +546,12 @@ Paging (HAP).=0A ### =
hvm\_port80=0A > `=3D <boolean>`=0A =0A+### highmem-start=0A+> `=3D =
<size>`=0A+=0A+Specify the memory boundary past which memory will be =
treated as highmem (x86=0A+debug hypervisor only).=0A+=0A ### idle\_latency=
\_factor=0A > `=3D <integer>`=0A =0A--- a/xen/arch/x86/domain_page.c=0A+++ =
b/xen/arch/x86/domain_page.c=0A@@ -66,8 +66,10 @@ void *map_domain_page(uns=
igned long mfn)=0A     struct mapcache_vcpu *vcache;=0A     struct =
vcpu_maphash_entry *hashent;=0A =0A+#ifdef NDEBUG=0A     if ( mfn <=3D =
PFN_DOWN(__pa(HYPERVISOR_VIRT_END - 1)) )=0A         return mfn_to_virt(mfn=
);=0A+#endif=0A =0A     v =3D mapcache_current_vcpu();=0A     if ( !v || =
is_hvm_vcpu(v) )=0A@@ -249,8 +251,10 @@ int mapcache_domain_init(struct =
domain *=0A     if ( is_hvm_domain(d) || is_idle_domain(d) )=0A         =
return 0;=0A =0A+#ifdef NDEBUG=0A     if ( !mem_hotplug && max_page <=3D =
PFN_DOWN(__pa(HYPERVISOR_VIRT_END - 1)) )=0A         return 0;=0A+#endif=0A=
 =0A     dcache->l1tab =3D xzalloc_array(l1_pgentry_t *, MAPCACHE_L2_ENTRIE=
S + 1);=0A     d->arch.perdomain_l2_pg[MAPCACHE_SLOT] =3D alloc_domheap_pag=
e(NULL, memf);=0A@@ -418,8 +422,10 @@ void *map_domain_page_global(unsigned=
 lo=0A =0A     ASSERT(!in_irq() && local_irq_is_enabled());=0A =0A+#ifdef =
NDEBUG=0A     if ( mfn <=3D PFN_DOWN(__pa(HYPERVISOR_VIRT_END - 1)) )=0A   =
      return mfn_to_virt(mfn);=0A+#endif=0A =0A     spin_lock(&globalmap_lo=
ck);=0A =0A--- a/xen/arch/x86/mm.c=0A+++ b/xen/arch/x86/mm.c=0A@@ -250,6 =
+250,14 @@ void __init init_frametable(void)=0A         init_spagetable();=
=0A }=0A =0A+#ifndef NDEBUG=0A+static unsigned int __read_mostly root_pgt_p=
v_xen_slots=0A+    =3D ROOT_PAGETABLE_PV_XEN_SLOTS;=0A+static l4_pgentry_t =
__read_mostly split_l4e;=0A+#else=0A+#define root_pgt_pv_xen_slots =
ROOT_PAGETABLE_PV_XEN_SLOTS=0A+#endif=0A+=0A void __init arch_init_memory(v=
oid)=0A {=0A     unsigned long i, pfn, rstart_pfn, rend_pfn, iostart_pfn, =
ioend_pfn;=0A@@ -344,6 +352,40 @@ void __init arch_init_memory(void)=0A    =
 efi_init_memory();=0A =0A     mem_sharing_init();=0A+=0A+#ifndef =
NDEBUG=0A+    if ( highmem_start )=0A+    {=0A+        unsigned long =
split_va =3D (unsigned long)__va(highmem_start);=0A+=0A+        if ( =
split_va < HYPERVISOR_VIRT_END &&=0A+             split_va - 1 =3D=3D =
(unsigned long)__va(highmem_start - 1) )=0A+        {=0A+            =
root_pgt_pv_xen_slots =3D l4_table_offset(split_va) -=0A+                  =
                  ROOT_PAGETABLE_FIRST_XEN_SLOT;=0A+            ASSERT(root=
_pgt_pv_xen_slots < ROOT_PAGETABLE_PV_XEN_SLOTS);=0A+            if ( =
l4_table_offset(split_va) =3D=3D l4_table_offset(split_va - 1) )=0A+       =
     {=0A+                l3_pgentry_t *l3tab =3D alloc_xen_pagetable();=0A=
+=0A+                if ( l3tab )=0A+                {=0A+                 =
   const l3_pgentry_t *l3idle =3D=0A+                        l4e_to_l3e(idl=
e_pg_table[l4_table_offset(split_va)]);=0A+=0A+                    for ( i =
=3D 0; i < l3_table_offset(split_va); ++i )=0A+                        =
l3tab[i] =3D l3idle[i];=0A+                    for ( ; i <=3D L3_PAGETABLE_=
ENTRIES; ++i )=0A+                        l3tab[i] =3D l3e_empty();=0A+    =
                split_l4e =3D l4e_from_pfn(virt_to_mfn(l3tab),=0A+         =
                                    __PAGE_HYPERVISOR);=0A+                =
}=0A+                else=0A+                    ++root_pgt_pv_xen_slots;=
=0A+            }=0A+        }=0A+    }=0A+#endif=0A }=0A =0A int =
page_is_ram_type(unsigned long mfn, unsigned long mem_type)=0A@@ -1320,7 =
+1362,12 @@ void init_guest_l4_table(l4_pgentry_t l4=0A     /* Xen private =
mappings. */=0A     memcpy(&l4tab[ROOT_PAGETABLE_FIRST_XEN_SLOT],=0A       =
     &idle_pg_table[ROOT_PAGETABLE_FIRST_XEN_SLOT],=0A-           =
ROOT_PAGETABLE_PV_XEN_SLOTS * sizeof(l4_pgentry_t));=0A+           =
root_pgt_pv_xen_slots * sizeof(l4_pgentry_t));=0A+#ifndef NDEBUG=0A+    if =
( l4e_get_intpte(split_l4e) )=0A+        l4tab[ROOT_PAGETABLE_FIRST_XEN_SLO=
T + root_pgt_pv_xen_slots] =3D=0A+            split_l4e;=0A+#endif=0A     =
l4tab[l4_table_offset(LINEAR_PT_VIRT_START)] =3D=0A         l4e_from_pfn(do=
main_page_map_to_mfn(l4tab), __PAGE_HYPERVISOR);=0A     l4tab[l4_table_offs=
et(PERDOMAIN_VIRT_START)] =3D=0A--- a/xen/arch/x86/setup.c=0A+++ b/xen/arch=
/x86/setup.c=0A@@ -82,6 +82,11 @@ boolean_param("noapic", skip_ioapic_setu=
=0A s8 __read_mostly xen_cpuidle =3D -1;=0A boolean_param("cpuidle", =
xen_cpuidle);=0A =0A+#ifndef NDEBUG=0A+unsigned long __initdata highmem_sta=
rt;=0A+size_param("highmem-start", highmem_start);=0A+#endif=0A+=0A =
cpumask_t __read_mostly cpu_present_map;=0A =0A unsigned long __read_mostly=
 xen_phys_start;=0A@@ -787,6 +792,14 @@ void __init __start_xen(unsigned =
long mb=0A     modules_headroom =3D bzimage_headroom(bootstrap_map(mod), =
mod->mod_end);=0A     bootstrap_map(NULL);=0A =0A+#ifndef highmem_start=0A+=
    /* Don't allow split below 4Gb. */=0A+    if ( highmem_start < GB(4) =
)=0A+        highmem_start =3D 0;=0A+    else /* align to L3 entry =
boundary */=0A+        highmem_start &=3D ~((1UL << L3_PAGETABLE_SHIFT) - =
1);=0A+#endif=0A+=0A     for ( i =3D boot_e820.nr_map-1; i >=3D 0; i-- =
)=0A     {=0A         uint64_t s, e, mask =3D (1UL << L2_PAGETABLE_SHIFT) =
- 1;=0A@@ -915,6 +928,9 @@ void __init __start_xen(unsigned long mb=0A     =
        /* Don't overlap with other modules. */=0A             end =3D =
consider_modules(s, e, size, mod, mbi->mods_count, j);=0A =0A+            =
if ( highmem_start && end > highmem_start )=0A+                continue;=0A=
+=0A             if ( s < end &&=0A                  (headroom ||=0A       =
            ((end - size) >> PAGE_SHIFT) > mod[j].mod_start) )=0A@@ -956,6 =
+972,8 @@ void __init __start_xen(unsigned long mb=0A     kexec_reserve_are=
a(&boot_e820);=0A =0A     setup_max_pdx();=0A+    if ( highmem_start )=0A+ =
       xenheap_max_mfn(PFN_DOWN(highmem_start));=0A =0A     /*=0A      * =
Walk every RAM region and map it in its entirety (on x86/64, at least)=0A@@=
 -1127,7 +1145,8 @@ void __init __start_xen(unsigned long mb=0A         =
unsigned long limit =3D virt_to_mfn(HYPERVISOR_VIRT_END - 1);=0A         =
uint64_t mask =3D PAGE_SIZE - 1;=0A =0A-        xenheap_max_mfn(limit);=0A+=
        if ( !highmem_start )=0A+            xenheap_max_mfn(limit);=0A =
=0A         /* Pass the remaining memory to the allocator. */=0A         =
for ( i =3D 0; i < boot_e820.nr_map; i++ )=0A--- a/xen/common/page_alloc.c=
=0A+++ b/xen/common/page_alloc.c=0A@@ -45,6 +45,7 @@=0A #include <asm/flush=
tlb.h>=0A #ifdef CONFIG_X86=0A #include <asm/p2m.h>=0A+#include <asm/setup.=
h> /* for highmem_start only */=0A #else=0A #define p2m_pod_offline_or_brok=
en_hit(pg) 0=0A #define p2m_pod_offline_or_broken_replace(pg) BUG_ON(pg =
!=3D NULL)=0A@@ -203,6 +204,25 @@ unsigned long __init alloc_boot_pages(=0A=
         pg =3D (r->e - nr_pfns) & ~(pfn_align - 1);=0A         if ( pg < =
r->s )=0A             continue;=0A+=0A+#if defined(CONFIG_X86) && =
!defined(NDEBUG)=0A+        /*=0A+         * Filtering pfn_align =3D=3D 1 =
since the only allocations using a bigger=0A+         * alignment are the =
ones used for setting up the frame table chunks.=0A+         * Those =
allocations get remapped anyway, i.e. them not having 1:1=0A+         * =
mappings always accessible is not a problem.=0A+         */=0A+        if =
( highmem_start && pfn_align =3D=3D 1 &&=0A+             r->e > PFN_DOWN(hi=
ghmem_start) )=0A+        {=0A+            pg =3D r->s;=0A+            if =
( pg + nr_pfns > PFN_DOWN(highmem_start) )=0A+                continue;=0A+=
            r->s =3D pg + nr_pfns;=0A+            return pg;=0A+        =
}=0A+#endif=0A+=0A         _e =3D r->e;=0A         r->e =3D pg;=0A         =
bootmem_region_add(pg + nr_pfns, _e);=0A--- a/xen/include/asm-x86/setup.h=
=0A+++ b/xen/include/asm-x86/setup.h=0A@@ -43,4 +43,10 @@ void microcode_gr=
ab_module(=0A =0A extern uint8_t kbd_shift_flags;=0A =0A+#ifdef NDEBUG=0A+#=
 define highmem_start 0=0A+#else=0A+extern unsigned long highmem_start;=0A+=
#endif=0A+=0A #endif=0A
--=__Part30019204.0__=
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

--=__Part30019204.0__=--