From: Wei Liu <wei.liu2@citrix.com>
To: Xen-devel <xen-devel@lists.xenproject.org>
Cc: George Dunlap <george.dunlap@eu.citrix.com>,
Andrew Cooper <andrew.cooper3@citrix.com>,
Wei Liu <wei.liu2@citrix.com>, Jan Beulich <JBeulich@suse.com>
Subject: [PATCH v4 26/31] x86/mm: move pv_{alloc, free}_page_type
Date: Thu, 17 Aug 2017 15:44:51 +0100 [thread overview]
Message-ID: <20170817144456.18989-27-wei.liu2@citrix.com> (raw)
In-Reply-To: <20170817144456.18989-1-wei.liu2@citrix.com>
Move them and the helper functions to pv/mm.c. Use bool in the moved
code where appropriate.
Put BUG() in the stub because the callers will call BUG or BUG_ON
anyway.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
---
v4: add BUG() in stubs
---
xen/arch/x86/domain.c | 1 +
xen/arch/x86/mm.c | 492 --------------------------------------------
xen/arch/x86/pv/mm.c | 491 +++++++++++++++++++++++++++++++++++++++++++
xen/include/asm-x86/mm.h | 3 -
xen/include/asm-x86/pv/mm.h | 14 ++
5 files changed, 506 insertions(+), 495 deletions(-)
diff --git a/xen/arch/x86/domain.c b/xen/arch/x86/domain.c
index 8fae38485c..e79a7de7e4 100644
--- a/xen/arch/x86/domain.c
+++ b/xen/arch/x86/domain.c
@@ -64,6 +64,7 @@
#include <compat/vcpu.h>
#include <asm/psr.h>
#include <asm/pv/domain.h>
+#include <asm/pv/mm.h>
#include <asm/pv/processor.h>
DEFINE_PER_CPU(struct vcpu *, curr_vcpu);
diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c
index 590e7ae65b..204c20d6fd 100644
--- a/xen/arch/x86/mm.c
+++ b/xen/arch/x86/mm.c
@@ -510,21 +510,6 @@ static void invalidate_shadow_ldt(struct vcpu *v, int flush)
}
-static int alloc_segdesc_page(struct page_info *page)
-{
- const struct domain *owner = page_get_owner(page);
- struct desc_struct *descs = __map_domain_page(page);
- unsigned i;
-
- for ( i = 0; i < 512; i++ )
- if ( unlikely(!check_descriptor(owner, &descs[i])) )
- break;
-
- unmap_domain_page(descs);
-
- return i == 512 ? 0 : -EINVAL;
-}
-
bool get_page_from_mfn(mfn_t mfn, struct domain *d)
{
struct page_info *page = mfn_to_page(mfn_x(mfn));
@@ -539,7 +524,6 @@ bool get_page_from_mfn(mfn_t mfn, struct domain *d)
return true;
}
-
int get_page_and_type_from_mfn(mfn_t mfn, unsigned long type, struct domain *d,
int partial, bool preemptible)
{
@@ -1165,172 +1149,6 @@ int put_page_from_l4e(l4_pgentry_t l4e, unsigned long pfn, int partial,
return 1;
}
-static int alloc_l1_table(struct page_info *page)
-{
- struct domain *d = page_get_owner(page);
- unsigned long pfn = page_to_mfn(page);
- l1_pgentry_t *pl1e;
- unsigned int i;
- int ret = 0;
-
- pl1e = map_domain_page(_mfn(pfn));
-
- for ( i = 0; i < L1_PAGETABLE_ENTRIES; i++ )
- {
- switch ( ret = get_page_from_l1e(pl1e[i], d, d) )
- {
- default:
- goto fail;
- case 0:
- break;
- case _PAGE_RW ... _PAGE_RW | PAGE_CACHE_ATTRS:
- ASSERT(!(ret & ~(_PAGE_RW | PAGE_CACHE_ATTRS)));
- l1e_flip_flags(pl1e[i], ret);
- break;
- }
-
- adjust_guest_l1e(pl1e[i], d);
- }
-
- unmap_domain_page(pl1e);
- return 0;
-
- fail:
- gdprintk(XENLOG_WARNING, "Failure in alloc_l1_table: slot %#x\n", i);
- while ( i-- > 0 )
- put_page_from_l1e(pl1e[i], d);
-
- unmap_domain_page(pl1e);
- return ret;
-}
-
-static int alloc_l2_table(struct page_info *page, unsigned long type,
- int preemptible)
-{
- struct domain *d = page_get_owner(page);
- unsigned long pfn = page_to_mfn(page);
- l2_pgentry_t *pl2e;
- unsigned int i;
- int rc = 0;
-
- pl2e = map_domain_page(_mfn(pfn));
-
- for ( i = page->nr_validated_ptes; i < L2_PAGETABLE_ENTRIES; i++ )
- {
- if ( preemptible && i > page->nr_validated_ptes
- && hypercall_preempt_check() )
- {
- page->nr_validated_ptes = i;
- rc = -ERESTART;
- break;
- }
-
- if ( !is_guest_l2_slot(d, type, i) ||
- (rc = get_page_from_l2e(pl2e[i], pfn, d)) > 0 )
- continue;
-
- if ( rc < 0 )
- {
- gdprintk(XENLOG_WARNING, "Failure in alloc_l2_table: slot %#x\n", i);
- while ( i-- > 0 )
- if ( is_guest_l2_slot(d, type, i) )
- put_page_from_l2e(pl2e[i], pfn);
- break;
- }
-
- adjust_guest_l2e(pl2e[i], d);
- }
-
- if ( rc >= 0 && (type & PGT_pae_xen_l2) )
- {
- /* Xen private mappings. */
- memcpy(&pl2e[COMPAT_L2_PAGETABLE_FIRST_XEN_SLOT(d)],
- &compat_idle_pg_table_l2[
- l2_table_offset(HIRO_COMPAT_MPT_VIRT_START)],
- COMPAT_L2_PAGETABLE_XEN_SLOTS(d) * sizeof(*pl2e));
- }
-
- unmap_domain_page(pl2e);
- return rc > 0 ? 0 : rc;
-}
-
-static int alloc_l3_table(struct page_info *page)
-{
- struct domain *d = page_get_owner(page);
- unsigned long pfn = page_to_mfn(page);
- l3_pgentry_t *pl3e;
- unsigned int i;
- int rc = 0, partial = page->partial_pte;
-
- pl3e = map_domain_page(_mfn(pfn));
-
- /*
- * PAE guests allocate full pages, but aren't required to initialize
- * more than the first four entries; when running in compatibility
- * mode, however, the full page is visible to the MMU, and hence all
- * 512 entries must be valid/verified, which is most easily achieved
- * by clearing them out.
- */
- if ( is_pv_32bit_domain(d) )
- memset(pl3e + 4, 0, (L3_PAGETABLE_ENTRIES - 4) * sizeof(*pl3e));
-
- for ( i = page->nr_validated_ptes; i < L3_PAGETABLE_ENTRIES;
- i++, partial = 0 )
- {
- if ( is_pv_32bit_domain(d) && (i == 3) )
- {
- if ( !(l3e_get_flags(pl3e[i]) & _PAGE_PRESENT) ||
- (l3e_get_flags(pl3e[i]) & l3_disallow_mask(d)) )
- rc = -EINVAL;
- else
- rc = get_page_and_type_from_mfn(
- _mfn(l3e_get_pfn(pl3e[i])),
- PGT_l2_page_table | PGT_pae_xen_l2, d, partial, true);
- }
- else if ( !is_guest_l3_slot(i) ||
- (rc = get_page_from_l3e(pl3e[i], pfn, d, partial)) > 0 )
- continue;
-
- if ( rc == -ERESTART )
- {
- page->nr_validated_ptes = i;
- page->partial_pte = partial ?: 1;
- }
- else if ( rc == -EINTR && i )
- {
- page->nr_validated_ptes = i;
- page->partial_pte = 0;
- rc = -ERESTART;
- }
- if ( rc < 0 )
- break;
-
- adjust_guest_l3e(pl3e[i], d);
- }
-
- if ( rc >= 0 && !pv_create_pae_xen_mappings(d, pl3e) )
- rc = -EINVAL;
- if ( rc < 0 && rc != -ERESTART && rc != -EINTR )
- {
- gdprintk(XENLOG_WARNING, "Failure in alloc_l3_table: slot %#x\n", i);
- if ( i )
- {
- page->nr_validated_ptes = i;
- page->partial_pte = 0;
- current->arch.old_guest_table = page;
- }
- while ( i-- > 0 )
- {
- if ( !is_guest_l3_slot(i) )
- continue;
- unadjust_guest_l3e(pl3e[i], d);
- }
- }
-
- unmap_domain_page(pl3e);
- return rc > 0 ? 0 : rc;
-}
-
bool fill_ro_mpt(unsigned long mfn)
{
l4_pgentry_t *l4tab = map_domain_page(_mfn(mfn));
@@ -1355,188 +1173,6 @@ void zap_ro_mpt(unsigned long mfn)
unmap_domain_page(l4tab);
}
-static int alloc_l4_table(struct page_info *page)
-{
- struct domain *d = page_get_owner(page);
- unsigned long pfn = page_to_mfn(page);
- l4_pgentry_t *pl4e = map_domain_page(_mfn(pfn));
- unsigned int i;
- int rc = 0, partial = page->partial_pte;
-
- for ( i = page->nr_validated_ptes; i < L4_PAGETABLE_ENTRIES;
- i++, partial = 0 )
- {
- if ( !is_guest_l4_slot(d, i) ||
- (rc = get_page_from_l4e(pl4e[i], pfn, d, partial)) > 0 )
- continue;
-
- if ( rc == -ERESTART )
- {
- page->nr_validated_ptes = i;
- page->partial_pte = partial ?: 1;
- }
- else if ( rc < 0 )
- {
- if ( rc != -EINTR )
- gdprintk(XENLOG_WARNING,
- "Failure in alloc_l4_table: slot %#x\n", i);
- if ( i )
- {
- page->nr_validated_ptes = i;
- page->partial_pte = 0;
- if ( rc == -EINTR )
- rc = -ERESTART;
- else
- {
- if ( current->arch.old_guest_table )
- page->nr_validated_ptes++;
- current->arch.old_guest_table = page;
- }
- }
- }
- if ( rc < 0 )
- {
- unmap_domain_page(pl4e);
- return rc;
- }
-
- adjust_guest_l4e(pl4e[i], d);
- }
-
- if ( rc >= 0 )
- {
- pv_init_guest_l4_table(pl4e, d, !VM_ASSIST(d, m2p_strict));
- atomic_inc(&d->arch.pv_domain.nr_l4_pages);
- rc = 0;
- }
- unmap_domain_page(pl4e);
-
- return rc;
-}
-
-static void free_l1_table(struct page_info *page)
-{
- struct domain *d = page_get_owner(page);
- unsigned long pfn = page_to_mfn(page);
- l1_pgentry_t *pl1e;
- unsigned int i;
-
- pl1e = map_domain_page(_mfn(pfn));
-
- for ( i = 0; i < L1_PAGETABLE_ENTRIES; i++ )
- put_page_from_l1e(pl1e[i], d);
-
- unmap_domain_page(pl1e);
-}
-
-
-static int free_l2_table(struct page_info *page, int preemptible)
-{
- struct domain *d = page_get_owner(page);
- unsigned long pfn = page_to_mfn(page);
- l2_pgentry_t *pl2e;
- unsigned int i = page->nr_validated_ptes - 1;
- int err = 0;
-
- pl2e = map_domain_page(_mfn(pfn));
-
- ASSERT(page->nr_validated_ptes);
- do {
- if ( is_guest_l2_slot(d, page->u.inuse.type_info, i) &&
- put_page_from_l2e(pl2e[i], pfn) == 0 &&
- preemptible && i && hypercall_preempt_check() )
- {
- page->nr_validated_ptes = i;
- err = -ERESTART;
- }
- } while ( !err && i-- );
-
- unmap_domain_page(pl2e);
-
- if ( !err )
- page->u.inuse.type_info &= ~PGT_pae_xen_l2;
-
- return err;
-}
-
-static int free_l3_table(struct page_info *page)
-{
- struct domain *d = page_get_owner(page);
- unsigned long pfn = page_to_mfn(page);
- l3_pgentry_t *pl3e;
- int rc = 0, partial = page->partial_pte;
- unsigned int i = page->nr_validated_ptes - !partial;
-
- pl3e = map_domain_page(_mfn(pfn));
-
- do {
- if ( is_guest_l3_slot(i) )
- {
- rc = put_page_from_l3e(pl3e[i], pfn, partial, 0);
- if ( rc < 0 )
- break;
- partial = 0;
- if ( rc > 0 )
- continue;
- unadjust_guest_l3e(pl3e[i], d);
- }
- } while ( i-- );
-
- unmap_domain_page(pl3e);
-
- if ( rc == -ERESTART )
- {
- page->nr_validated_ptes = i;
- page->partial_pte = partial ?: -1;
- }
- else if ( rc == -EINTR && i < L3_PAGETABLE_ENTRIES - 1 )
- {
- page->nr_validated_ptes = i + 1;
- page->partial_pte = 0;
- rc = -ERESTART;
- }
- return rc > 0 ? 0 : rc;
-}
-
-static int free_l4_table(struct page_info *page)
-{
- struct domain *d = page_get_owner(page);
- unsigned long pfn = page_to_mfn(page);
- l4_pgentry_t *pl4e = map_domain_page(_mfn(pfn));
- int rc = 0, partial = page->partial_pte;
- unsigned int i = page->nr_validated_ptes - !partial;
-
- do {
- if ( is_guest_l4_slot(d, i) )
- rc = put_page_from_l4e(pl4e[i], pfn, partial, 0);
- if ( rc < 0 )
- break;
- partial = 0;
- } while ( i-- );
-
- if ( rc == -ERESTART )
- {
- page->nr_validated_ptes = i;
- page->partial_pte = partial ?: -1;
- }
- else if ( rc == -EINTR && i < L4_PAGETABLE_ENTRIES - 1 )
- {
- page->nr_validated_ptes = i + 1;
- page->partial_pte = 0;
- rc = -ERESTART;
- }
-
- unmap_domain_page(pl4e);
-
- if ( rc >= 0 )
- {
- atomic_dec(&d->arch.pv_domain.nr_l4_pages);
- rc = 0;
- }
-
- return rc;
-}
-
int page_lock(struct page_info *page)
{
unsigned long x, nx;
@@ -1951,134 +1587,6 @@ void get_page_light(struct page_info *page)
while ( unlikely(y != x) );
}
-static int pv_alloc_page_type(struct page_info *page, unsigned long type,
- bool preemptible)
-{
- struct domain *owner = page_get_owner(page);
- int rc;
-
- /* A page table is dirtied when its type count becomes non-zero. */
- if ( likely(owner != NULL) )
- paging_mark_dirty(owner, _mfn(page_to_mfn(page)));
-
- switch ( type & PGT_type_mask )
- {
- case PGT_l1_page_table:
- rc = alloc_l1_table(page);
- break;
- case PGT_l2_page_table:
- rc = alloc_l2_table(page, type, preemptible);
- break;
- case PGT_l3_page_table:
- ASSERT(preemptible);
- rc = alloc_l3_table(page);
- break;
- case PGT_l4_page_table:
- ASSERT(preemptible);
- rc = alloc_l4_table(page);
- break;
- case PGT_seg_desc_page:
- rc = alloc_segdesc_page(page);
- break;
- default:
- printk("Bad type in alloc_page_type %lx t=%" PRtype_info " c=%lx\n",
- type, page->u.inuse.type_info,
- page->count_info);
- rc = -EINVAL;
- BUG();
- }
-
- /* No need for atomic update of type_info here: noone else updates it. */
- smp_wmb();
- switch ( rc )
- {
- case 0:
- page->u.inuse.type_info |= PGT_validated;
- break;
- case -EINTR:
- ASSERT((page->u.inuse.type_info &
- (PGT_count_mask|PGT_validated|PGT_partial)) == 1);
- page->u.inuse.type_info &= ~PGT_count_mask;
- break;
- default:
- ASSERT(rc < 0);
- gdprintk(XENLOG_WARNING, "Error while validating mfn %" PRI_mfn
- " (pfn %" PRI_pfn ") for type %" PRtype_info
- ": caf=%08lx taf=%" PRtype_info "\n",
- page_to_mfn(page), get_gpfn_from_mfn(page_to_mfn(page)),
- type, page->count_info, page->u.inuse.type_info);
- if ( page != current->arch.old_guest_table )
- page->u.inuse.type_info = 0;
- else
- {
- ASSERT((page->u.inuse.type_info &
- (PGT_count_mask | PGT_validated)) == 1);
- case -ERESTART:
- get_page_light(page);
- page->u.inuse.type_info |= PGT_partial;
- }
- break;
- }
-
- return rc;
-}
-
-
-int pv_free_page_type(struct page_info *page, unsigned long type,
- bool preemptible)
-{
- struct domain *owner = page_get_owner(page);
- unsigned long gmfn;
- int rc;
-
- if ( likely(owner != NULL) && unlikely(paging_mode_enabled(owner)) )
- {
- /* A page table is dirtied when its type count becomes zero. */
- paging_mark_dirty(owner, _mfn(page_to_mfn(page)));
-
- ASSERT(!shadow_mode_refcounts(owner));
-
- gmfn = mfn_to_gmfn(owner, page_to_mfn(page));
- ASSERT(VALID_M2P(gmfn));
- /* Page sharing not supported for shadowed domains */
- if(!SHARED_M2P(gmfn))
- shadow_remove_all_shadows(owner, _mfn(gmfn));
- }
-
- if ( !(type & PGT_partial) )
- {
- page->nr_validated_ptes = 1U << PAGETABLE_ORDER;
- page->partial_pte = 0;
- }
-
- switch ( type & PGT_type_mask )
- {
- case PGT_l1_page_table:
- free_l1_table(page);
- rc = 0;
- break;
- case PGT_l2_page_table:
- rc = free_l2_table(page, preemptible);
- break;
- case PGT_l3_page_table:
- ASSERT(preemptible);
- rc = free_l3_table(page);
- break;
- case PGT_l4_page_table:
- ASSERT(preemptible);
- rc = free_l4_table(page);
- break;
- default:
- gdprintk(XENLOG_WARNING, "type %" PRtype_info " mfn %" PRI_mfn "\n",
- type, page_to_mfn(page));
- rc = -EINVAL;
- BUG();
- }
-
- return rc;
-}
-
-
static int __put_final_page_type(
struct page_info *page, unsigned long type, int preemptible)
{
diff --git a/xen/arch/x86/pv/mm.c b/xen/arch/x86/pv/mm.c
index 46e1fcf4e5..f0393b9e3c 100644
--- a/xen/arch/x86/pv/mm.c
+++ b/xen/arch/x86/pv/mm.c
@@ -20,10 +20,13 @@
* along with this program; If not, see <http://www.gnu.org/licenses/>.
*/
+#include <xen/event.h>
#include <xen/guest_access.h>
+#include <asm/mm.h>
#include <asm/pv/mm.h>
#include <asm/setup.h>
+#include <asm/shadow.h>
/*
* PTE updates can be done with ordinary writes except:
@@ -251,6 +254,494 @@ bool pv_create_pae_xen_mappings(struct domain *d, l3_pgentry_t *pl3e)
return true;
}
+static int alloc_l1_table(struct page_info *page)
+{
+ struct domain *d = page_get_owner(page);
+ unsigned long pfn = page_to_mfn(page);
+ l1_pgentry_t *pl1e;
+ unsigned int i;
+ int ret = 0;
+
+ pl1e = map_domain_page(_mfn(pfn));
+
+ for ( i = 0; i < L1_PAGETABLE_ENTRIES; i++ )
+ {
+ switch ( ret = get_page_from_l1e(pl1e[i], d, d) )
+ {
+ default:
+ goto fail;
+ case 0:
+ break;
+ case _PAGE_RW ... _PAGE_RW | PAGE_CACHE_ATTRS:
+ ASSERT(!(ret & ~(_PAGE_RW | PAGE_CACHE_ATTRS)));
+ l1e_flip_flags(pl1e[i], ret);
+ break;
+ }
+
+ adjust_guest_l1e(pl1e[i], d);
+ }
+
+ unmap_domain_page(pl1e);
+ return 0;
+
+ fail:
+ gdprintk(XENLOG_WARNING, "Failure in alloc_l1_table: slot %#x\n", i);
+ while ( i-- > 0 )
+ put_page_from_l1e(pl1e[i], d);
+
+ unmap_domain_page(pl1e);
+ return ret;
+}
+
+static int alloc_l2_table(struct page_info *page, unsigned long type,
+ bool preemptible)
+{
+ struct domain *d = page_get_owner(page);
+ unsigned long pfn = page_to_mfn(page);
+ l2_pgentry_t *pl2e;
+ unsigned int i;
+ int rc = 0;
+
+ pl2e = map_domain_page(_mfn(pfn));
+
+ for ( i = page->nr_validated_ptes; i < L2_PAGETABLE_ENTRIES; i++ )
+ {
+ if ( preemptible && i > page->nr_validated_ptes
+ && hypercall_preempt_check() )
+ {
+ page->nr_validated_ptes = i;
+ rc = -ERESTART;
+ break;
+ }
+
+ if ( !is_guest_l2_slot(d, type, i) ||
+ (rc = get_page_from_l2e(pl2e[i], pfn, d)) > 0 )
+ continue;
+
+ if ( rc < 0 )
+ {
+ gdprintk(XENLOG_WARNING, "Failure in alloc_l2_table: slot %#x\n", i);
+ while ( i-- > 0 )
+ if ( is_guest_l2_slot(d, type, i) )
+ put_page_from_l2e(pl2e[i], pfn);
+ break;
+ }
+
+ adjust_guest_l2e(pl2e[i], d);
+ }
+
+ if ( rc >= 0 && (type & PGT_pae_xen_l2) )
+ {
+ /* Xen private mappings. */
+ memcpy(&pl2e[COMPAT_L2_PAGETABLE_FIRST_XEN_SLOT(d)],
+ &compat_idle_pg_table_l2[
+ l2_table_offset(HIRO_COMPAT_MPT_VIRT_START)],
+ COMPAT_L2_PAGETABLE_XEN_SLOTS(d) * sizeof(*pl2e));
+ }
+
+ unmap_domain_page(pl2e);
+ return rc > 0 ? 0 : rc;
+}
+
+static int alloc_l3_table(struct page_info *page)
+{
+ struct domain *d = page_get_owner(page);
+ unsigned long pfn = page_to_mfn(page);
+ l3_pgentry_t *pl3e;
+ unsigned int i;
+ int rc = 0, partial = page->partial_pte;
+
+ pl3e = map_domain_page(_mfn(pfn));
+
+ /*
+ * PAE guests allocate full pages, but aren't required to initialize
+ * more than the first four entries; when running in compatibility
+ * mode, however, the full page is visible to the MMU, and hence all
+ * 512 entries must be valid/verified, which is most easily achieved
+ * by clearing them out.
+ */
+ if ( is_pv_32bit_domain(d) )
+ memset(pl3e + 4, 0, (L3_PAGETABLE_ENTRIES - 4) * sizeof(*pl3e));
+
+ for ( i = page->nr_validated_ptes; i < L3_PAGETABLE_ENTRIES;
+ i++, partial = 0 )
+ {
+ if ( is_pv_32bit_domain(d) && (i == 3) )
+ {
+ if ( !(l3e_get_flags(pl3e[i]) & _PAGE_PRESENT) ||
+ (l3e_get_flags(pl3e[i]) & l3_disallow_mask(d)) )
+ rc = -EINVAL;
+ else
+ rc = get_page_and_type_from_mfn(
+ _mfn(l3e_get_pfn(pl3e[i])),
+ PGT_l2_page_table | PGT_pae_xen_l2, d, partial, true);
+ }
+ else if ( !is_guest_l3_slot(i) ||
+ (rc = get_page_from_l3e(pl3e[i], pfn, d, partial)) > 0 )
+ continue;
+
+ if ( rc == -ERESTART )
+ {
+ page->nr_validated_ptes = i;
+ page->partial_pte = partial ?: 1;
+ }
+ else if ( rc == -EINTR && i )
+ {
+ page->nr_validated_ptes = i;
+ page->partial_pte = 0;
+ rc = -ERESTART;
+ }
+ if ( rc < 0 )
+ break;
+
+ adjust_guest_l3e(pl3e[i], d);
+ }
+
+ if ( rc >= 0 && !pv_create_pae_xen_mappings(d, pl3e) )
+ rc = -EINVAL;
+ if ( rc < 0 && rc != -ERESTART && rc != -EINTR )
+ {
+ gdprintk(XENLOG_WARNING, "Failure in alloc_l3_table: slot %#x\n", i);
+ if ( i )
+ {
+ page->nr_validated_ptes = i;
+ page->partial_pte = 0;
+ current->arch.old_guest_table = page;
+ }
+ while ( i-- > 0 )
+ {
+ if ( !is_guest_l3_slot(i) )
+ continue;
+ unadjust_guest_l3e(pl3e[i], d);
+ }
+ }
+
+ unmap_domain_page(pl3e);
+ return rc > 0 ? 0 : rc;
+}
+
+static int alloc_l4_table(struct page_info *page)
+{
+ struct domain *d = page_get_owner(page);
+ unsigned long pfn = page_to_mfn(page);
+ l4_pgentry_t *pl4e = map_domain_page(_mfn(pfn));
+ unsigned int i;
+ int rc = 0, partial = page->partial_pte;
+
+ for ( i = page->nr_validated_ptes; i < L4_PAGETABLE_ENTRIES;
+ i++, partial = 0 )
+ {
+ if ( !is_guest_l4_slot(d, i) ||
+ (rc = get_page_from_l4e(pl4e[i], pfn, d, partial)) > 0 )
+ continue;
+
+ if ( rc == -ERESTART )
+ {
+ page->nr_validated_ptes = i;
+ page->partial_pte = partial ?: 1;
+ }
+ else if ( rc < 0 )
+ {
+ if ( rc != -EINTR )
+ gdprintk(XENLOG_WARNING,
+ "Failure in alloc_l4_table: slot %#x\n", i);
+ if ( i )
+ {
+ page->nr_validated_ptes = i;
+ page->partial_pte = 0;
+ if ( rc == -EINTR )
+ rc = -ERESTART;
+ else
+ {
+ if ( current->arch.old_guest_table )
+ page->nr_validated_ptes++;
+ current->arch.old_guest_table = page;
+ }
+ }
+ }
+ if ( rc < 0 )
+ {
+ unmap_domain_page(pl4e);
+ return rc;
+ }
+
+ adjust_guest_l4e(pl4e[i], d);
+ }
+
+ if ( rc >= 0 )
+ {
+ pv_init_guest_l4_table(pl4e, d, !VM_ASSIST(d, m2p_strict));
+ atomic_inc(&d->arch.pv_domain.nr_l4_pages);
+ rc = 0;
+ }
+ unmap_domain_page(pl4e);
+
+ return rc;
+}
+
+static void free_l1_table(struct page_info *page)
+{
+ struct domain *d = page_get_owner(page);
+ unsigned long pfn = page_to_mfn(page);
+ l1_pgentry_t *pl1e;
+ unsigned int i;
+
+ pl1e = map_domain_page(_mfn(pfn));
+
+ for ( i = 0; i < L1_PAGETABLE_ENTRIES; i++ )
+ put_page_from_l1e(pl1e[i], d);
+
+ unmap_domain_page(pl1e);
+}
+
+static int free_l2_table(struct page_info *page, int preemptible)
+{
+ struct domain *d = page_get_owner(page);
+ unsigned long pfn = page_to_mfn(page);
+ l2_pgentry_t *pl2e;
+ unsigned int i = page->nr_validated_ptes - 1;
+ int err = 0;
+
+ pl2e = map_domain_page(_mfn(pfn));
+
+ ASSERT(page->nr_validated_ptes);
+ do {
+ if ( is_guest_l2_slot(d, page->u.inuse.type_info, i) &&
+ put_page_from_l2e(pl2e[i], pfn) == 0 &&
+ preemptible && i && hypercall_preempt_check() )
+ {
+ page->nr_validated_ptes = i;
+ err = -ERESTART;
+ }
+ } while ( !err && i-- );
+
+ unmap_domain_page(pl2e);
+
+ if ( !err )
+ page->u.inuse.type_info &= ~PGT_pae_xen_l2;
+
+ return err;
+}
+
+static int free_l3_table(struct page_info *page)
+{
+ struct domain *d = page_get_owner(page);
+ unsigned long pfn = page_to_mfn(page);
+ l3_pgentry_t *pl3e;
+ int rc = 0, partial = page->partial_pte;
+ unsigned int i = page->nr_validated_ptes - !partial;
+
+ pl3e = map_domain_page(_mfn(pfn));
+
+ do {
+ if ( is_guest_l3_slot(i) )
+ {
+ rc = put_page_from_l3e(pl3e[i], pfn, partial, 0);
+ if ( rc < 0 )
+ break;
+ partial = 0;
+ if ( rc > 0 )
+ continue;
+ unadjust_guest_l3e(pl3e[i], d);
+ }
+ } while ( i-- );
+
+ unmap_domain_page(pl3e);
+
+ if ( rc == -ERESTART )
+ {
+ page->nr_validated_ptes = i;
+ page->partial_pte = partial ?: -1;
+ }
+ else if ( rc == -EINTR && i < L3_PAGETABLE_ENTRIES - 1 )
+ {
+ page->nr_validated_ptes = i + 1;
+ page->partial_pte = 0;
+ rc = -ERESTART;
+ }
+ return rc > 0 ? 0 : rc;
+}
+
+static int free_l4_table(struct page_info *page)
+{
+ struct domain *d = page_get_owner(page);
+ unsigned long pfn = page_to_mfn(page);
+ l4_pgentry_t *pl4e = map_domain_page(_mfn(pfn));
+ int rc = 0, partial = page->partial_pte;
+ unsigned int i = page->nr_validated_ptes - !partial;
+
+ do {
+ if ( is_guest_l4_slot(d, i) )
+ rc = put_page_from_l4e(pl4e[i], pfn, partial, 0);
+ if ( rc < 0 )
+ break;
+ partial = 0;
+ } while ( i-- );
+
+ if ( rc == -ERESTART )
+ {
+ page->nr_validated_ptes = i;
+ page->partial_pte = partial ?: -1;
+ }
+ else if ( rc == -EINTR && i < L4_PAGETABLE_ENTRIES - 1 )
+ {
+ page->nr_validated_ptes = i + 1;
+ page->partial_pte = 0;
+ rc = -ERESTART;
+ }
+
+ unmap_domain_page(pl4e);
+
+ if ( rc >= 0 )
+ {
+ atomic_dec(&d->arch.pv_domain.nr_l4_pages);
+ rc = 0;
+ }
+
+ return rc;
+}
+
+static int alloc_segdesc_page(struct page_info *page)
+{
+ const struct domain *owner = page_get_owner(page);
+ struct desc_struct *descs = __map_domain_page(page);
+ unsigned i;
+
+ for ( i = 0; i < 512; i++ )
+ if ( unlikely(!check_descriptor(owner, &descs[i])) )
+ break;
+
+ unmap_domain_page(descs);
+
+ return i == 512 ? 0 : -EINVAL;
+}
+
+int pv_alloc_page_type(struct page_info *page, unsigned long type,
+ bool preemptible)
+{
+ struct domain *owner = page_get_owner(page);
+ int rc;
+
+ /* A page table is dirtied when its type count becomes non-zero. */
+ if ( likely(owner != NULL) )
+ paging_mark_dirty(owner, _mfn(page_to_mfn(page)));
+
+ switch ( type & PGT_type_mask )
+ {
+ case PGT_l1_page_table:
+ rc = alloc_l1_table(page);
+ break;
+ case PGT_l2_page_table:
+ rc = alloc_l2_table(page, type, preemptible);
+ break;
+ case PGT_l3_page_table:
+ ASSERT(preemptible);
+ rc = alloc_l3_table(page);
+ break;
+ case PGT_l4_page_table:
+ ASSERT(preemptible);
+ rc = alloc_l4_table(page);
+ break;
+ case PGT_seg_desc_page:
+ rc = alloc_segdesc_page(page);
+ break;
+ default:
+ printk("Bad type in alloc_page_type %lx t=%" PRtype_info " c=%lx\n",
+ type, page->u.inuse.type_info,
+ page->count_info);
+ rc = -EINVAL;
+ BUG();
+ }
+
+ /* No need for atomic update of type_info here: noone else updates it. */
+ smp_wmb();
+ switch ( rc )
+ {
+ case 0:
+ page->u.inuse.type_info |= PGT_validated;
+ break;
+ case -EINTR:
+ ASSERT((page->u.inuse.type_info &
+ (PGT_count_mask|PGT_validated|PGT_partial)) == 1);
+ page->u.inuse.type_info &= ~PGT_count_mask;
+ break;
+ default:
+ ASSERT(rc < 0);
+ gdprintk(XENLOG_WARNING, "Error while validating mfn %" PRI_mfn
+ " (pfn %" PRI_pfn ") for type %" PRtype_info
+ ": caf=%08lx taf=%" PRtype_info "\n",
+ page_to_mfn(page), get_gpfn_from_mfn(page_to_mfn(page)),
+ type, page->count_info, page->u.inuse.type_info);
+ if ( page != current->arch.old_guest_table )
+ page->u.inuse.type_info = 0;
+ else
+ {
+ ASSERT((page->u.inuse.type_info &
+ (PGT_count_mask | PGT_validated)) == 1);
+ case -ERESTART:
+ get_page_light(page);
+ page->u.inuse.type_info |= PGT_partial;
+ }
+ break;
+ }
+
+ return rc;
+}
+
+int pv_free_page_type(struct page_info *page, unsigned long type,
+ bool preemptible)
+{
+ struct domain *owner = page_get_owner(page);
+ unsigned long gmfn;
+ int rc;
+
+ if ( likely(owner != NULL) && unlikely(paging_mode_enabled(owner)) )
+ {
+ /* A page table is dirtied when its type count becomes zero. */
+ paging_mark_dirty(owner, _mfn(page_to_mfn(page)));
+
+ ASSERT(!shadow_mode_refcounts(owner));
+
+ gmfn = mfn_to_gmfn(owner, page_to_mfn(page));
+ ASSERT(VALID_M2P(gmfn));
+ /* Page sharing not supported for shadowed domains */
+ if(!SHARED_M2P(gmfn))
+ shadow_remove_all_shadows(owner, _mfn(gmfn));
+ }
+
+ if ( !(type & PGT_partial) )
+ {
+ page->nr_validated_ptes = 1U << PAGETABLE_ORDER;
+ page->partial_pte = 0;
+ }
+
+ switch ( type & PGT_type_mask )
+ {
+ case PGT_l1_page_table:
+ free_l1_table(page);
+ rc = 0;
+ break;
+ case PGT_l2_page_table:
+ rc = free_l2_table(page, preemptible);
+ break;
+ case PGT_l3_page_table:
+ ASSERT(preemptible);
+ rc = free_l3_table(page);
+ break;
+ case PGT_l4_page_table:
+ ASSERT(preemptible);
+ rc = free_l4_table(page);
+ break;
+ default:
+ gdprintk(XENLOG_WARNING, "type %" PRtype_info " mfn %" PRI_mfn "\n",
+ type, page_to_mfn(page));
+ rc = -EINVAL;
+ BUG();
+ }
+
+ return rc;
+}
+
/*
* Local variables:
* mode: C
diff --git a/xen/include/asm-x86/mm.h b/xen/include/asm-x86/mm.h
index 6857651db1..7480341240 100644
--- a/xen/include/asm-x86/mm.h
+++ b/xen/include/asm-x86/mm.h
@@ -302,9 +302,6 @@ static inline void *__page_to_virt(const struct page_info *pg)
(PAGE_SIZE / (sizeof(*pg) & -sizeof(*pg))));
}
-int pv_free_page_type(struct page_info *page, unsigned long type,
- bool preemptible);
-
bool_t fill_ro_mpt(unsigned long mfn);
void zap_ro_mpt(unsigned long mfn);
diff --git a/xen/include/asm-x86/pv/mm.h b/xen/include/asm-x86/pv/mm.h
index 648b26d7d0..7de5ea2f12 100644
--- a/xen/include/asm-x86/pv/mm.h
+++ b/xen/include/asm-x86/pv/mm.h
@@ -95,10 +95,17 @@ void pv_arch_init_memory(void);
int pv_new_guest_cr3(unsigned long pfn);
+int pv_alloc_page_type(struct page_info *page, unsigned long type,
+ bool preemptible);
+int pv_free_page_type(struct page_info *page, unsigned long type,
+ bool preemptible);
+
#else
#include <xen/errno.h>
+#include <asm/bug.h>
+
static inline void pv_get_guest_eff_l1e(unsigned long addr,
l1_pgentry_t *eff_l1e)
{}
@@ -125,6 +132,13 @@ static inline void pv_arch_init_memory(void) {}
static inline int pv_new_guest_cr3(unsigned long pfn) { return -EINVAL; }
+static inline int pv_alloc_page_type(struct page_info *page, unsigned long type,
+ bool preemptible)
+{ BUG(); return -EINVAL; }
+static inline int pv_free_page_type(struct page_info *page, unsigned long type,
+ bool preemptible)
+{ BUG(); return -EINVAL; }
+
#endif
#endif /* __X86_PV_MM_H__ */
--
2.11.0
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel
next prev parent reply other threads:[~2017-08-17 14:47 UTC|newest]
Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-08-17 14:44 [PATCH v4 00/31] x86: refactor mm.c Wei Liu
2017-08-17 14:44 ` [PATCH v4 01/31] x86/mm: carve out create_grant_pv_mapping Wei Liu
2017-08-18 10:12 ` Jan Beulich
2017-08-17 14:44 ` [PATCH v4 02/31] x86/mm: carve out replace_grant_pv_mapping Wei Liu
2017-08-18 10:14 ` Jan Beulich
2017-08-17 14:44 ` [PATCH v4 03/31] x86/mm: split HVM grant table code to hvm/grant_table.c Wei Liu
2017-08-18 10:16 ` Jan Beulich
2017-08-18 10:26 ` Andrew Cooper
2017-08-17 14:44 ` [PATCH v4 04/31] x86/mm: lift PAGE_CACHE_ATTRS to page.h Wei Liu
2017-08-17 14:46 ` Andrew Cooper
2017-08-17 14:44 ` [PATCH v4 05/31] x86/mm: document the return values from get_page_from_l*e Wei Liu
2017-08-18 10:24 ` Jan Beulich
2017-08-17 14:44 ` [PATCH v4 06/31] x86: move pv_emul_is_mem_write to pv/emulate.c Wei Liu
2017-08-17 14:53 ` Andrew Cooper
2017-08-18 10:08 ` Jan Beulich
2017-08-18 12:08 ` Wei Liu
2017-08-18 12:13 ` Andrew Cooper
2017-08-17 14:44 ` [PATCH v4 07/31] x86/mm: move and rename guest_get_eff{, kern}_l1e Wei Liu
2017-08-24 14:52 ` Jan Beulich
2017-08-17 14:44 ` [PATCH v4 08/31] x86/mm: export get_page_from_mfn Wei Liu
2017-08-24 14:55 ` Jan Beulich
2017-08-17 14:44 ` [PATCH v4 09/31] x86/mm: rename and move update_intpte Wei Liu
2017-08-24 14:59 ` Jan Beulich
2017-08-17 14:44 ` [PATCH v4 10/31] x86/mm: move {un, }adjust_guest_* to pv/mm.h Wei Liu
2017-08-24 15:00 ` Jan Beulich
2017-08-17 14:44 ` [PATCH v4 11/31] x86/mm: split out writable pagetable emulation code Wei Liu
2017-08-24 15:15 ` Jan Beulich
2017-08-30 14:07 ` Wei Liu
2017-08-30 15:23 ` Jan Beulich
2017-08-30 15:43 ` Wei Liu
2017-08-17 14:44 ` [PATCH v4 12/31] x86/mm: split out readonly MMIO " Wei Liu
2017-08-24 15:16 ` Jan Beulich
2017-08-24 15:25 ` Andrew Cooper
2017-08-30 14:35 ` Wei Liu
2017-08-17 14:44 ` [PATCH v4 13/31] x86/mm: remove the unused inclusion of pv/emulate.h Wei Liu
2017-08-17 14:44 ` [PATCH v4 14/31] x86/mm: move and rename guest_{, un}map_l1e Wei Liu
2017-08-17 14:44 ` [PATCH v4 15/31] x86/mm: split out PV grant table code Wei Liu
2017-08-17 14:44 ` [PATCH v4 16/31] x86/mm: split out descriptor " Wei Liu
2017-08-17 14:44 ` [PATCH v4 17/31] x86/mm: move compat descriptor handling code Wei Liu
2017-08-17 14:44 ` [PATCH v4 18/31] x86/mm: move and rename map_ldt_shadow_page Wei Liu
2017-08-17 14:44 ` [PATCH v4 19/31] x86/mm: factor out pv_arch_init_memory Wei Liu
2017-08-17 14:44 ` [PATCH v4 20/31] x86/mm: move l4 table setup code Wei Liu
2017-08-17 14:44 ` [PATCH v4 21/31] x86/mm: add "pv_" prefix to new_guest_cr3 Wei Liu
2017-08-17 14:44 ` [PATCH v4 22/31] x86: add pv_ prefix to {alloc, free}_page_type Wei Liu
2017-08-17 14:44 ` [PATCH v4 23/31] x86/mm: export more get/put page functions Wei Liu
2017-08-17 14:44 ` [PATCH v4 24/31] x86/mm: move and add pv_ prefix to create_pae_xen_mappings Wei Liu
2017-08-17 14:44 ` [PATCH v4 25/31] x86/mm: move disallow_mask variable and macros Wei Liu
2017-08-17 14:44 ` Wei Liu [this message]
2017-08-17 14:44 ` [PATCH v4 27/31] x86/mm: move and add pv_ prefix to invalidate_shadow_ldt Wei Liu
2017-08-17 14:44 ` [PATCH v4 28/31] x86/mm: move PV hypercalls to pv/mm-hypercalls.c Wei Liu
2017-08-17 14:44 ` [PATCH v4 29/31] x86/mm: remove the now unused inclusion of pv/mm.h Wei Liu
2017-08-17 14:44 ` [PATCH v4 30/31] x86/mm: use put_page_type_preemptible in put_page_from_l{2, 3}e Wei Liu
2017-08-17 14:44 ` [PATCH v4 31/31] x86/mm: move {get, put}_page_from_l{2, 3, 4}e Wei Liu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170817144456.18989-27-wei.liu2@citrix.com \
--to=wei.liu2@citrix.com \
--cc=JBeulich@suse.com \
--cc=andrew.cooper3@citrix.com \
--cc=george.dunlap@eu.citrix.com \
--cc=xen-devel@lists.xenproject.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.