[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-changelog] [xen master] x86/shadow: compile most write emulation code just once



commit 2c80710a78c931457ed1215d19834efe26ff115d
Author:     Jan Beulich <jbeulich@xxxxxxxx>
AuthorDate: Thu Mar 10 16:48:23 2016 +0100
Commit:     Jan Beulich <jbeulich@xxxxxxxx>
CommitDate: Thu Mar 10 16:48:23 2016 +0100

    x86/shadow: compile most write emulation code just once
    
    No need to compile all of this code three times, as most of it really
    is guest mode independent. The savings are between 3k and 4k of binary
    code in my builds. For this to fully work out, the sh_gva_to_gfn()
    calls are being replaced by indirect ones through the paging mode
    table.
    
    No functional change (i.e. only formatting and naming changes) except
    for
    - sh_emulate_map_dest()'s user mode check corrected for the PV case
      (affecting debugging mode only, this isn't being split out)
    - simplifying the vaddr argument to emulate_gva_to_mfn() for the second
      part in the cross page write case
    
    Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
    Reviewed-by: Tim Deegan <tim@xxxxxxx>
    Reviewed-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
---
 xen/arch/x86/mm/shadow/common.c  | 205 +++++++++++++++++++++++++++++++++++++++
 xen/arch/x86/mm/shadow/multi.c   | 197 ++-----------------------------------
 xen/arch/x86/mm/shadow/private.h |  11 +++
 3 files changed, 222 insertions(+), 191 deletions(-)

diff --git a/xen/arch/x86/mm/shadow/common.c b/xen/arch/x86/mm/shadow/common.c
index cdad572..fc1628b 100644
--- a/xen/arch/x86/mm/shadow/common.c
+++ b/xen/arch/x86/mm/shadow/common.c
@@ -1679,6 +1679,211 @@ static unsigned int shadow_get_allocation(struct domain 
*d)
 }
 
 /**************************************************************************/
+/* Handling guest writes to pagetables. */
+
+/* Translate a VA to an MFN, injecting a page-fault if we fail. */
+#define BAD_GVA_TO_GFN (~0UL)
+#define BAD_GFN_TO_MFN (~1UL)
+#define READONLY_GFN   (~2UL)
+static mfn_t emulate_gva_to_mfn(struct vcpu *v, unsigned long vaddr,
+                                struct sh_emulate_ctxt *sh_ctxt)
+{
+    unsigned long gfn;
+    struct page_info *page;
+    mfn_t mfn;
+    p2m_type_t p2mt;
+    uint32_t pfec = PFEC_page_present | PFEC_write_access;
+
+    /* Translate the VA to a GFN. */
+    gfn = paging_get_hostmode(v)->gva_to_gfn(v, NULL, vaddr, &pfec);
+    if ( gfn == INVALID_GFN )
+    {
+        if ( is_hvm_vcpu(v) )
+            hvm_inject_page_fault(pfec, vaddr);
+        else
+            propagate_page_fault(vaddr, pfec);
+        return _mfn(BAD_GVA_TO_GFN);
+    }
+
+    /* Translate the GFN to an MFN. */
+    ASSERT(!paging_locked_by_me(v->domain));
+
+    page = get_page_from_gfn(v->domain, gfn, &p2mt, P2M_ALLOC);
+
+    /* Sanity checking. */
+    if ( page == NULL )
+    {
+        return _mfn(BAD_GFN_TO_MFN);
+    }
+    if ( p2m_is_discard_write(p2mt) )
+    {
+        put_page(page);
+        return _mfn(READONLY_GFN);
+    }
+    if ( !p2m_is_ram(p2mt) )
+    {
+        put_page(page);
+        return _mfn(BAD_GFN_TO_MFN);
+    }
+    mfn = page_to_mfn(page);
+    ASSERT(mfn_valid(mfn));
+
+    v->arch.paging.last_write_was_pt = !!sh_mfn_is_a_page_table(mfn);
+    /*
+     * Note shadow cannot page out or unshare this mfn, so the map won't
+     * disappear. Otherwise, caller must hold onto page until done.
+     */
+    put_page(page);
+
+    return mfn;
+}
+
+/* Check that the user is allowed to perform this write. */
+void *sh_emulate_map_dest(struct vcpu *v, unsigned long vaddr,
+                          unsigned int bytes,
+                          struct sh_emulate_ctxt *sh_ctxt)
+{
+    struct domain *d = v->domain;
+    void *map;
+
+    sh_ctxt->mfn1 = emulate_gva_to_mfn(v, vaddr, sh_ctxt);
+    if ( !mfn_valid(sh_ctxt->mfn1) )
+        return ((mfn_x(sh_ctxt->mfn1) == BAD_GVA_TO_GFN) ?
+                MAPPING_EXCEPTION :
+                (mfn_x(sh_ctxt->mfn1) == READONLY_GFN) ?
+                MAPPING_SILENT_FAIL : MAPPING_UNHANDLEABLE);
+
+#ifndef NDEBUG
+    /* We don't emulate user-mode writes to page tables. */
+    if ( has_hvm_container_domain(d)
+         ? hvm_get_seg_reg(x86_seg_ss, sh_ctxt)->attr.fields.dpl == 3
+         : !guest_kernel_mode(v, guest_cpu_user_regs()) )
+    {
+        gdprintk(XENLOG_DEBUG, "User-mode write to pagetable reached "
+                 "emulate_map_dest(). This should never happen!\n");
+        return MAPPING_UNHANDLEABLE;
+    }
+#endif
+
+    /* Unaligned writes mean probably this isn't a pagetable. */
+    if ( vaddr & (bytes - 1) )
+        sh_remove_shadows(d, sh_ctxt->mfn1, 0, 0 /* Slow, can fail. */ );
+
+    if ( likely(((vaddr + bytes - 1) & PAGE_MASK) == (vaddr & PAGE_MASK)) )
+    {
+        /* Whole write fits on a single page. */
+        sh_ctxt->mfn2 = _mfn(INVALID_MFN);
+        map = map_domain_page(sh_ctxt->mfn1) + (vaddr & ~PAGE_MASK);
+    }
+    else
+    {
+        mfn_t mfns[2];
+
+        /*
+         * Cross-page emulated writes are only supported for HVM guests;
+         * PV guests ought to know better.
+         */
+        if ( !is_hvm_domain(d) )
+            return MAPPING_UNHANDLEABLE;
+
+        /* This write crosses a page boundary. Translate the second page. */
+        sh_ctxt->mfn2 = emulate_gva_to_mfn(v, vaddr + bytes - 1, sh_ctxt);
+        if ( !mfn_valid(sh_ctxt->mfn2) )
+            return ((mfn_x(sh_ctxt->mfn2) == BAD_GVA_TO_GFN) ?
+                    MAPPING_EXCEPTION :
+                    (mfn_x(sh_ctxt->mfn2) == READONLY_GFN) ?
+                    MAPPING_SILENT_FAIL : MAPPING_UNHANDLEABLE);
+
+        /* Cross-page writes mean probably not a pagetable. */
+        sh_remove_shadows(d, sh_ctxt->mfn2, 0, 0 /* Slow, can fail. */ );
+
+        mfns[0] = sh_ctxt->mfn1;
+        mfns[1] = sh_ctxt->mfn2;
+        map = vmap(mfns, 2);
+        if ( !map )
+            return MAPPING_UNHANDLEABLE;
+        map += (vaddr & ~PAGE_MASK);
+    }
+
+#if (SHADOW_OPTIMIZATIONS & SHOPT_SKIP_VERIFY)
+    /*
+     * Remember if the bottom bit was clear, so we can choose not to run
+     * the change through the verify code if it's still clear afterwards.
+     */
+    sh_ctxt->low_bit_was_clear = map != NULL && !(*(u8 *)map & _PAGE_PRESENT);
+#endif
+
+    return map;
+}
+
+/*
+ * Tidy up after the emulated write: mark pages dirty, verify the new
+ * contents, and undo the mapping.
+ */
+void sh_emulate_unmap_dest(struct vcpu *v, void *addr, unsigned int bytes,
+                           struct sh_emulate_ctxt *sh_ctxt)
+{
+    u32 b1 = bytes, b2 = 0, shflags;
+
+    /*
+     * We can avoid re-verifying the page contents after the write if:
+     *  - it was no larger than the PTE type of this pagetable;
+     *  - it was aligned to the PTE boundaries; and
+     *  - _PAGE_PRESENT was clear before and after the write.
+     */
+    shflags = mfn_to_page(sh_ctxt->mfn1)->shadow_flags;
+#if (SHADOW_OPTIMIZATIONS & SHOPT_SKIP_VERIFY)
+    if ( sh_ctxt->low_bit_was_clear
+         && !(*(u8 *)addr & _PAGE_PRESENT)
+         && ((!(shflags & SHF_32)
+              /*
+               * Not shadowed 32-bit: aligned 64-bit writes that leave
+               * the present bit unset are safe to ignore.
+               */
+              && ((unsigned long)addr & 7) == 0
+              && bytes <= 8)
+             ||
+             (!(shflags & (SHF_PAE|SHF_64))
+              /*
+               * Not shadowed PAE/64-bit: aligned 32-bit writes that
+               * leave the present bit unset are safe to ignore.
+               */
+              && ((unsigned long)addr & 3) == 0
+              && bytes <= 4)) )
+    {
+        /* Writes with this alignment constraint can't possibly cross pages. */
+        ASSERT(!mfn_valid(sh_ctxt->mfn2));
+    }
+    else
+#endif /* SHADOW_OPTIMIZATIONS & SHOPT_SKIP_VERIFY */
+    {
+        if ( unlikely(mfn_valid(sh_ctxt->mfn2)) )
+        {
+            /* Validate as two writes, one to each page. */
+            b1 = PAGE_SIZE - (((unsigned long)addr) & ~PAGE_MASK);
+            b2 = bytes - b1;
+            ASSERT(b2 < bytes);
+        }
+        if ( likely(b1 > 0) )
+            sh_validate_guest_pt_write(v, sh_ctxt->mfn1, addr, b1);
+        if ( unlikely(b2 > 0) )
+            sh_validate_guest_pt_write(v, sh_ctxt->mfn2, addr + b1, b2);
+    }
+
+    paging_mark_dirty(v->domain, mfn_x(sh_ctxt->mfn1));
+
+    if ( unlikely(mfn_valid(sh_ctxt->mfn2)) )
+    {
+        paging_mark_dirty(v->domain, mfn_x(sh_ctxt->mfn2));
+        vunmap((void *)((unsigned long)addr & PAGE_MASK));
+    }
+    else
+        unmap_domain_page(addr);
+
+    atomic_inc(&v->domain->arch.paging.shadow.gtable_dirty_version);
+}
+
+/**************************************************************************/
 /* Hash table for storing the guest->shadow mappings.
  * The table itself is an array of pointers to shadows; the shadows are then
  * threaded on a singly-linked list of shadows with the same hash value */
diff --git a/xen/arch/x86/mm/shadow/multi.c b/xen/arch/x86/mm/shadow/multi.c
index 01d916d..c334779 100644
--- a/xen/arch/x86/mm/shadow/multi.c
+++ b/xen/arch/x86/mm/shadow/multi.c
@@ -4606,141 +4606,7 @@ static void sh_pagetable_dying(struct vcpu *v, paddr_t 
gpa)
 #endif
 
 /**************************************************************************/
-/* Handling HVM guest writes to pagetables  */
-
-/* Translate a VA to an MFN, injecting a page-fault if we fail */
-#define BAD_GVA_TO_GFN (~0UL)
-#define BAD_GFN_TO_MFN (~1UL)
-#define READONLY_GFN   (~2UL)
-static mfn_t emulate_gva_to_mfn(struct vcpu *v,
-                                unsigned long vaddr,
-                                struct sh_emulate_ctxt *sh_ctxt)
-{
-    unsigned long gfn;
-    struct page_info *page;
-    mfn_t mfn;
-    p2m_type_t p2mt;
-    uint32_t pfec = PFEC_page_present | PFEC_write_access;
-
-    /* Translate the VA to a GFN */
-    gfn = sh_gva_to_gfn(v, NULL, vaddr, &pfec);
-    if ( gfn == INVALID_GFN )
-    {
-        if ( is_hvm_vcpu(v) )
-            hvm_inject_page_fault(pfec, vaddr);
-        else
-            propagate_page_fault(vaddr, pfec);
-        return _mfn(BAD_GVA_TO_GFN);
-    }
-
-    /* Translate the GFN to an MFN */
-    ASSERT(!paging_locked_by_me(v->domain));
-
-    page = get_page_from_gfn(v->domain, gfn, &p2mt, P2M_ALLOC);
-
-    /* Sanity checking */
-    if ( page == NULL )
-    {
-        return _mfn(BAD_GFN_TO_MFN);
-    }
-    if ( p2m_is_discard_write(p2mt) )
-    {
-        put_page(page);
-        return _mfn(READONLY_GFN);
-    }
-    if ( !p2m_is_ram(p2mt) )
-    {
-        put_page(page);
-        return _mfn(BAD_GFN_TO_MFN);
-    }
-    mfn = page_to_mfn(page);
-    ASSERT(mfn_valid(mfn));
-
-    v->arch.paging.last_write_was_pt = !!sh_mfn_is_a_page_table(mfn);
-    /* Note shadow cannot page out or unshare this mfn, so the map won't
-     * disappear. Otherwise, caller must hold onto page until done. */
-    put_page(page);
-    return mfn;
-}
-
-/* Check that the user is allowed to perform this write.
- * Returns a mapped pointer to write to, or NULL for error. */
-#define MAPPING_UNHANDLEABLE ((void *)(unsigned long)X86EMUL_UNHANDLEABLE)
-#define MAPPING_EXCEPTION    ((void *)(unsigned long)X86EMUL_EXCEPTION)
-#define MAPPING_SILENT_FAIL  ((void *)(unsigned long)X86EMUL_OKAY)
-#define emulate_map_dest_failed(rc) ((unsigned long)(rc) <= 3)
-static void *emulate_map_dest(struct vcpu *v,
-                              unsigned long vaddr,
-                              u32 bytes,
-                              struct sh_emulate_ctxt *sh_ctxt)
-{
-    struct domain *d = v->domain;
-    void *map = NULL;
-
-    sh_ctxt->mfn1 = emulate_gva_to_mfn(v, vaddr, sh_ctxt);
-    if ( !mfn_valid(sh_ctxt->mfn1) )
-        return ((mfn_x(sh_ctxt->mfn1) == BAD_GVA_TO_GFN) ?
-                MAPPING_EXCEPTION :
-                (mfn_x(sh_ctxt->mfn1) == READONLY_GFN) ?
-                MAPPING_SILENT_FAIL : MAPPING_UNHANDLEABLE);
-
-#ifndef NDEBUG
-    /* We don't emulate user-mode writes to page tables */
-    if ( hvm_get_seg_reg(x86_seg_ss, sh_ctxt)->attr.fields.dpl == 3 )
-    {
-        gdprintk(XENLOG_DEBUG, "User-mode write to pagetable reached "
-                 "emulate_map_dest(). This should never happen!\n");
-        return MAPPING_UNHANDLEABLE;
-    }
-#endif
-
-    /* Unaligned writes mean probably this isn't a pagetable */
-    if ( vaddr & (bytes - 1) )
-        sh_remove_shadows(d, sh_ctxt->mfn1, 0, 0 /* Slow, can fail */ );
-
-    if ( likely(((vaddr + bytes - 1) & PAGE_MASK) == (vaddr & PAGE_MASK)) )
-    {
-        /* Whole write fits on a single page */
-        sh_ctxt->mfn2 = _mfn(INVALID_MFN);
-        map = map_domain_page(sh_ctxt->mfn1) + (vaddr & ~PAGE_MASK);
-    }
-    else
-    {
-        mfn_t mfns[2];
-
-        /* Cross-page emulated writes are only supported for HVM guests;
-         * PV guests ought to know better */
-        if ( !is_hvm_domain(d) )
-            return MAPPING_UNHANDLEABLE;
-
-        /* This write crosses a page boundary.  Translate the second page */
-        sh_ctxt->mfn2 = emulate_gva_to_mfn(v, (vaddr + bytes - 1) & PAGE_MASK,
-                                           sh_ctxt);
-        if ( !mfn_valid(sh_ctxt->mfn2) )
-            return ((mfn_x(sh_ctxt->mfn2) == BAD_GVA_TO_GFN) ?
-                    MAPPING_EXCEPTION :
-                    (mfn_x(sh_ctxt->mfn2) == READONLY_GFN) ?
-                    MAPPING_SILENT_FAIL : MAPPING_UNHANDLEABLE);
-
-        /* Cross-page writes mean probably not a pagetable */
-        sh_remove_shadows(d, sh_ctxt->mfn2, 0, 0 /* Slow, can fail */ );
-
-        mfns[0] = sh_ctxt->mfn1;
-        mfns[1] = sh_ctxt->mfn2;
-        map = vmap(mfns, 2);
-        if ( !map )
-            return MAPPING_UNHANDLEABLE;
-        map += (vaddr & ~PAGE_MASK);
-    }
-
-#if (SHADOW_OPTIMIZATIONS & SHOPT_SKIP_VERIFY)
-    /* Remember if the bottom bit was clear, so we can choose not to run
-     * the change through the verify code if it's still clear afterwards */
-    sh_ctxt->low_bit_was_clear = map != NULL && !(*(u8 *)map & _PAGE_PRESENT);
-#endif
-
-    return map;
-}
+/* Handling guest writes to pagetables. */
 
 /* Tidy up after the emulated write: mark pages dirty, verify the new
  * contents, and undo the mapping */
@@ -4749,8 +4615,6 @@ static void emulate_unmap_dest(struct vcpu *v,
                                u32 bytes,
                                struct sh_emulate_ctxt *sh_ctxt)
 {
-    u32 b1 = bytes, b2 = 0, shflags;
-
     ASSERT(mfn_valid(sh_ctxt->mfn1));
 
     /* If we are writing lots of PTE-aligned zeros, might want to unshadow */
@@ -4764,56 +4628,7 @@ static void emulate_unmap_dest(struct vcpu *v,
     else
         reset_early_unshadow(v);
 
-    /* We can avoid re-verifying the page contents after the write if:
-     *  - it was no larger than the PTE type of this pagetable;
-     *  - it was aligned to the PTE boundaries; and
-     *  - _PAGE_PRESENT was clear before and after the write. */
-    shflags = mfn_to_page(sh_ctxt->mfn1)->shadow_flags;
-#if (SHADOW_OPTIMIZATIONS & SHOPT_SKIP_VERIFY)
-    if ( sh_ctxt->low_bit_was_clear
-         && !(*(u8 *)addr & _PAGE_PRESENT)
-         && ((!(shflags & SHF_32)
-              /* Not shadowed 32-bit: aligned 64-bit writes that leave
-               * the present bit unset are safe to ignore. */
-              && ((unsigned long)addr & 7) == 0
-              && bytes <= 8)
-             ||
-             (!(shflags & (SHF_PAE|SHF_64))
-              /* Not shadowed PAE/64-bit: aligned 32-bit writes that
-               * leave the present bit unset are safe to ignore. */
-              && ((unsigned long)addr & 3) == 0
-              && bytes <= 4)) )
-    {
-        /* Writes with this alignment constraint can't possibly cross pages */
-        ASSERT(!mfn_valid(sh_ctxt->mfn2));
-    }
-    else
-#endif /* SHADOW_OPTIMIZATIONS & SHOPT_SKIP_VERIFY */
-    {
-        if ( unlikely(mfn_valid(sh_ctxt->mfn2)) )
-        {
-            /* Validate as two writes, one to each page */
-            b1 = PAGE_SIZE - (((unsigned long)addr) & ~PAGE_MASK);
-            b2 = bytes - b1;
-            ASSERT(b2 < bytes);
-        }
-        if ( likely(b1 > 0) )
-            sh_validate_guest_pt_write(v, sh_ctxt->mfn1, addr, b1);
-        if ( unlikely(b2 > 0) )
-            sh_validate_guest_pt_write(v, sh_ctxt->mfn2, addr + b1, b2);
-    }
-
-    paging_mark_dirty(v->domain, mfn_x(sh_ctxt->mfn1));
-
-    if ( unlikely(mfn_valid(sh_ctxt->mfn2)) )
-    {
-        paging_mark_dirty(v->domain, mfn_x(sh_ctxt->mfn2));
-        vunmap((void *)((unsigned long)addr & PAGE_MASK));
-    }
-    else
-        unmap_domain_page(addr);
-
-    atomic_inc(&v->domain->arch.paging.shadow.gtable_dirty_version);
+    sh_emulate_unmap_dest(v, addr, bytes, sh_ctxt);
 }
 
 static int
@@ -4826,8 +4641,8 @@ sh_x86_emulate_write(struct vcpu *v, unsigned long vaddr, 
void *src,
     if ( (vaddr & (bytes - 1)) && !is_hvm_vcpu(v)  )
         return X86EMUL_UNHANDLEABLE;
 
-    addr = emulate_map_dest(v, vaddr, bytes, sh_ctxt);
-    if ( emulate_map_dest_failed(addr) )
+    addr = sh_emulate_map_dest(v, vaddr, bytes, sh_ctxt);
+    if ( sh_emulate_map_dest_failed(addr) )
         return (long)addr;
 
     paging_lock(v->domain);
@@ -4868,8 +4683,8 @@ sh_x86_emulate_cmpxchg(struct vcpu *v, unsigned long 
vaddr,
     if ( (vaddr & (bytes - 1)) && !is_hvm_vcpu(v)  )
         return X86EMUL_UNHANDLEABLE;
 
-    addr = emulate_map_dest(v, vaddr, bytes, sh_ctxt);
-    if ( emulate_map_dest_failed(addr) )
+    addr = sh_emulate_map_dest(v, vaddr, bytes, sh_ctxt);
+    if ( sh_emulate_map_dest_failed(addr) )
         return (long)addr;
 
     paging_lock(v->domain);
diff --git a/xen/arch/x86/mm/shadow/private.h b/xen/arch/x86/mm/shadow/private.h
index a881a4e..e061cbb 100644
--- a/xen/arch/x86/mm/shadow/private.h
+++ b/xen/arch/x86/mm/shadow/private.h
@@ -393,6 +393,17 @@ void shadow_update_paging_modes(struct vcpu *v);
  * With user_only == 1, unhooks only the user-mode mappings. */
 void shadow_unhook_mappings(struct domain *d, mfn_t smfn, int user_only);
 
+/* Returns a mapped pointer to write to, or one of the following error
+ * indicators. */
+#define MAPPING_UNHANDLEABLE ((void *)(unsigned long)X86EMUL_UNHANDLEABLE)
+#define MAPPING_EXCEPTION    ((void *)(unsigned long)X86EMUL_EXCEPTION)
+#define MAPPING_SILENT_FAIL  ((void *)(unsigned long)X86EMUL_OKAY)
+#define sh_emulate_map_dest_failed(rc) ((unsigned long)(rc) <= 3)
+void *sh_emulate_map_dest(struct vcpu *v, unsigned long vaddr,
+                          unsigned int bytes, struct sh_emulate_ctxt *sh_ctxt);
+void sh_emulate_unmap_dest(struct vcpu *v, void *addr, unsigned int bytes,
+                           struct sh_emulate_ctxt *sh_ctxt);
+
 #if (SHADOW_OPTIMIZATIONS & SHOPT_OUT_OF_SYNC)
 /* Allow a shadowed page to go out of sync */
 int sh_unsync(struct vcpu *v, mfn_t gmfn);
--
generated by git-patchbot for /home/xen/git/xen.git#master

_______________________________________________
Xen-changelog mailing list
Xen-changelog@xxxxxxxxxxxxx
http://lists.xensource.com/xen-changelog

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.