xen: virtual mmu

Xen pagetable handling, including the machinery to implement direct
pagetables.

Xen presents the real CPU's pagetables directly to guests, with no
added shadowing or other layer of abstraction.  Naturally this means
the hypervisor must maintain close control over what the guest can put
into the pagetable.

When the guest modifies the pte/pmd/pgd, it must convert its
domain-specific notion of a "physical" pfn into a global machine frame
number (mfn) before inserting the entry into the pagetable.  Xen will
check to make sure the domain is allowed to create a mapping of the
given mfn.

Xen also requires that all mappings the guest has of its own active
pagetable are read-only.  This is relatively easy to implement in
Linux because all pagetables share the same pte pages for kernel
mappings, so updating the pte in one pagetable will implicitly update
the mapping in all pagetables.

Normally a pagetable becomes active when you point to it with cr3 (or
the Xen equivalent), but when you do so, Xen must check the whole
pagetable for correctness, which is clearly a performance problem.

Xen solves this with pinning which keeps a pagetable effectively
active even if its currently unused, which means that all the normal
update rules are enforced.  This means that it need not revalidate the
pagetable when loading cr3.

This patch has a first-cut implementation of pinning, but it is more
fully implemented in a later patch.

Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>

arch/i386/xen/Makefile

@@ -1 +1 @@
-obj-y		:= enlighten.o setup.o features.o multicalls.o
+obj-y		:= enlighten.o setup.o features.o multicalls.o mmu.o

arch/i386/xen/enlighten.c

@@ -39,6 +39,7 @@
 #include <asm/pgtable.h>
 
 #include "xen-ops.h"
+#include "mmu.h"
 #include "multicalls.h"
 
 EXPORT_SYMBOL_GPL(hypercall_page);
@@ -579,11 +580,9 @@ static __init void xen_pagetable_setup_done(pgd_t *base)
 		 * Should be set_fixmap(), but shared_info is a machine
 		 * address with no corresponding pseudo-phys address.
 		 */
-#if 0
 		set_pte_mfn(fix_to_virt(FIX_PARAVIRT_BOOTMAP),
 			    PFN_DOWN(xen_start_info->shared_info),
 			    PAGE_KERNEL);
-#endif
 
 		HYPERVISOR_shared_info =
 			(struct shared_info *)fix_to_virt(FIX_PARAVIRT_BOOTMAP);
@@ -592,9 +591,7 @@ static __init void xen_pagetable_setup_done(pgd_t *base)
 		HYPERVISOR_shared_info =
 			(struct shared_info *)__va(xen_start_info->shared_info);
 
-#if 0
 	xen_pgd_pin(base);
-#endif
 
 	xen_vcpu_setup(smp_processor_id());
 }
@@ -690,6 +687,31 @@ static const struct paravirt_ops xen_paravirt_ops __initdata = {
 	.release_pd = xen_release_pd,
 	.release_pt = xen_release_pt,
 
+	.set_pte = xen_set_pte,
+	.set_pte_at = xen_set_pte_at,
+	.set_pmd = xen_set_pmd,
+
+	.pte_val = xen_pte_val,
+	.pgd_val = xen_pgd_val,
+
+	.make_pte = xen_make_pte,
+	.make_pgd = xen_make_pgd,
+
+#ifdef CONFIG_X86_PAE
+	.set_pte_atomic = xen_set_pte_atomic,
+	.set_pte_present = xen_set_pte_at,
+	.set_pud = xen_set_pud,
+	.pte_clear = xen_pte_clear,
+	.pmd_clear = xen_pmd_clear,
+
+	.make_pmd = xen_make_pmd,
+	.pmd_val = xen_pmd_val,
+#endif	/* PAE */
+
+	.activate_mm = xen_activate_mm,
+	.dup_mmap = xen_dup_mmap,
+	.exit_mmap = xen_exit_mmap,
+
 	.set_lazy_mode = xen_set_lazy_mode,
 };
 

arch/i386/xen/mmu.c

@@ -0,0 +1,420 @@
+/*
+ * Xen mmu operations
+ *
+ * This file contains the various mmu fetch and update operations.
+ * The most important job they must perform is the mapping between the
+ * domain's pfn and the overall machine mfns.
+ *
+ * Xen allows guests to directly update the pagetable, in a controlled
+ * fashion.  In other words, the guest modifies the same pagetable
+ * that the CPU actually uses, which eliminates the overhead of having
+ * a separate shadow pagetable.
+ *
+ * In order to allow this, it falls on the guest domain to map its
+ * notion of a "physical" pfn - which is just a domain-local linear
+ * address - into a real "machine address" which the CPU's MMU can
+ * use.
+ *
+ * A pgd_t/pmd_t/pte_t will typically contain an mfn, and so can be
+ * inserted directly into the pagetable.  When creating a new
+ * pte/pmd/pgd, it converts the passed pfn into an mfn.  Conversely,
+ * when reading the content back with __(pgd|pmd|pte)_val, it converts
+ * the mfn back into a pfn.
+ *
+ * The other constraint is that all pages which make up a pagetable
+ * must be mapped read-only in the guest.  This prevents uncontrolled
+ * guest updates to the pagetable.  Xen strictly enforces this, and
+ * will disallow any pagetable update which will end up mapping a
+ * pagetable page RW, and will disallow using any writable page as a
+ * pagetable.
+ *
+ * Naively, when loading %cr3 with the base of a new pagetable, Xen
+ * would need to validate the whole pagetable before going on.
+ * Naturally, this is quite slow.  The solution is to "pin" a
+ * pagetable, which enforces all the constraints on the pagetable even
+ * when it is not actively in use.  This menas that Xen can be assured
+ * that it is still valid when you do load it into %cr3, and doesn't
+ * need to revalidate it.
+ *
+ * Jeremy Fitzhardinge <jeremy@xensource.com>, XenSource Inc, 2007
+ */
+#include <linux/bug.h>
+#include <linux/sched.h>
+
+#include <asm/pgtable.h>
+#include <asm/tlbflush.h>
+#include <asm/mmu_context.h>
+
+#include <asm/xen/hypercall.h>
+#include <asm/paravirt.h>
+
+#include <xen/page.h>
+#include <xen/interface/xen.h>
+
+#include "mmu.h"
+
+xmaddr_t arbitrary_virt_to_machine(unsigned long address)
+{
+	pte_t *pte = lookup_address(address);
+	unsigned offset = address & PAGE_MASK;
+
+	BUG_ON(pte == NULL);
+
+	return XMADDR((pte_mfn(*pte) << PAGE_SHIFT) + offset);
+}
+
+void make_lowmem_page_readonly(void *vaddr)
+{
+	pte_t *pte, ptev;
+	unsigned long address = (unsigned long)vaddr;
+
+	pte = lookup_address(address);
+	BUG_ON(pte == NULL);
+
+	ptev = pte_wrprotect(*pte);
+
+	if (HYPERVISOR_update_va_mapping(address, ptev, 0))
+		BUG();
+}
+
+void make_lowmem_page_readwrite(void *vaddr)
+{
+	pte_t *pte, ptev;
+	unsigned long address = (unsigned long)vaddr;
+
+	pte = lookup_address(address);
+	BUG_ON(pte == NULL);
+
+	ptev = pte_mkwrite(*pte);
+
+	if (HYPERVISOR_update_va_mapping(address, ptev, 0))
+		BUG();
+}
+
+
+void xen_set_pte(pte_t *ptep, pte_t pte)
+{
+	struct mmu_update u;
+
+	u.ptr = virt_to_machine(ptep).maddr;
+	u.val = pte_val_ma(pte);
+	if (HYPERVISOR_mmu_update(&u, 1, NULL, DOMID_SELF) < 0)
+		BUG();
+}
+
+void xen_set_pmd(pmd_t *ptr, pmd_t val)
+{
+	struct mmu_update u;
+
+	u.ptr = virt_to_machine(ptr).maddr;
+	u.val = pmd_val_ma(val);
+	if (HYPERVISOR_mmu_update(&u, 1, NULL, DOMID_SELF) < 0)
+		BUG();
+}
+
+#ifdef CONFIG_X86_PAE
+void xen_set_pud(pud_t *ptr, pud_t val)
+{
+	struct mmu_update u;
+
+	u.ptr = virt_to_machine(ptr).maddr;
+	u.val = pud_val_ma(val);
+	if (HYPERVISOR_mmu_update(&u, 1, NULL, DOMID_SELF) < 0)
+		BUG();
+}
+#endif
+
+/*
+ * Associate a virtual page frame with a given physical page frame
+ * and protection flags for that frame.
+ */
+void set_pte_mfn(unsigned long vaddr, unsigned long mfn, pgprot_t flags)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+
+	pgd = swapper_pg_dir + pgd_index(vaddr);
+	if (pgd_none(*pgd)) {
+		BUG();
+		return;
+	}
+	pud = pud_offset(pgd, vaddr);
+	if (pud_none(*pud)) {
+		BUG();
+		return;
+	}
+	pmd = pmd_offset(pud, vaddr);
+	if (pmd_none(*pmd)) {
+		BUG();
+		return;
+	}
+	pte = pte_offset_kernel(pmd, vaddr);
+	/* <mfn,flags> stored as-is, to permit clearing entries */
+	xen_set_pte(pte, mfn_pte(mfn, flags));
+
+	/*
+	 * It's enough to flush this one mapping.
+	 * (PGE mappings get flushed as well)
+	 */
+	__flush_tlb_one(vaddr);
+}
+
+void xen_set_pte_at(struct mm_struct *mm, unsigned long addr,
+		    pte_t *ptep, pte_t pteval)
+{
+	if ((mm != current->mm && mm != &init_mm) ||
+	    HYPERVISOR_update_va_mapping(addr, pteval, 0) != 0)
+		xen_set_pte(ptep, pteval);
+}
+
+#ifdef CONFIG_X86_PAE
+void xen_set_pte_atomic(pte_t *ptep, pte_t pte)
+{
+	set_64bit((u64 *)ptep, pte_val_ma(pte));
+}
+
+void xen_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+{
+	ptep->pte_low = 0;
+	smp_wmb();		/* make sure low gets written first */
+	ptep->pte_high = 0;
+}
+
+void xen_pmd_clear(pmd_t *pmdp)
+{
+	xen_set_pmd(pmdp, __pmd(0));
+}
+
+unsigned long long xen_pte_val(pte_t pte)
+{
+	unsigned long long ret = 0;
+
+	if (pte.pte_low) {
+		ret = ((unsigned long long)pte.pte_high << 32) | pte.pte_low;
+		ret = machine_to_phys(XMADDR(ret)).paddr | 1;
+	}
+
+	return ret;
+}
+
+unsigned long long xen_pmd_val(pmd_t pmd)
+{
+	unsigned long long ret = pmd.pmd;
+	if (ret)
+		ret = machine_to_phys(XMADDR(ret)).paddr | 1;
+	return ret;
+}
+
+unsigned long long xen_pgd_val(pgd_t pgd)
+{
+	unsigned long long ret = pgd.pgd;
+	if (ret)
+		ret = machine_to_phys(XMADDR(ret)).paddr | 1;
+	return ret;
+}
+
+pte_t xen_make_pte(unsigned long long pte)
+{
+	if (pte & 1)
+		pte = phys_to_machine(XPADDR(pte)).maddr;
+
+	return (pte_t){ pte, pte >> 32 };
+}
+
+pmd_t xen_make_pmd(unsigned long long pmd)
+{
+	if (pmd & 1)
+		pmd = phys_to_machine(XPADDR(pmd)).maddr;
+
+	return (pmd_t){ pmd };
+}
+
+pgd_t xen_make_pgd(unsigned long long pgd)
+{
+	if (pgd & _PAGE_PRESENT)
+		pgd = phys_to_machine(XPADDR(pgd)).maddr;
+
+	return (pgd_t){ pgd };
+}
+#else  /* !PAE */
+unsigned long xen_pte_val(pte_t pte)
+{
+	unsigned long ret = pte.pte_low;
+
+	if (ret & _PAGE_PRESENT)
+		ret = machine_to_phys(XMADDR(ret)).paddr;
+
+	return ret;
+}
+
+unsigned long xen_pmd_val(pmd_t pmd)
+{
+	/* a BUG here is a lot easier to track down than a NULL eip */
+	BUG();
+	return 0;
+}
+
+unsigned long xen_pgd_val(pgd_t pgd)
+{
+	unsigned long ret = pgd.pgd;
+	if (ret)
+		ret = machine_to_phys(XMADDR(ret)).paddr | 1;
+	return ret;
+}
+
+pte_t xen_make_pte(unsigned long pte)
+{
+	if (pte & _PAGE_PRESENT)
+		pte = phys_to_machine(XPADDR(pte)).maddr;
+
+	return (pte_t){ pte };
+}
+
+pmd_t xen_make_pmd(unsigned long pmd)
+{
+	/* a BUG here is a lot easier to track down than a NULL eip */
+	BUG();
+	return __pmd(0);
+}
+
+pgd_t xen_make_pgd(unsigned long pgd)
+{
+	if (pgd & _PAGE_PRESENT)
+		pgd = phys_to_machine(XPADDR(pgd)).maddr;
+
+	return (pgd_t){ pgd };
+}
+#endif	/* CONFIG_X86_PAE */
+
+
+
+static void pgd_walk_set_prot(void *pt, pgprot_t flags)
+{
+	unsigned long pfn = PFN_DOWN(__pa(pt));
+
+	if (HYPERVISOR_update_va_mapping((unsigned long)pt,
+					 pfn_pte(pfn, flags), 0) < 0)
+		BUG();
+}
+
+static void pgd_walk(pgd_t *pgd_base, pgprot_t flags)
+{
+	pgd_t *pgd = pgd_base;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+	int    g, u, m;
+
+	if (xen_feature(XENFEAT_auto_translated_physmap))
+		return;
+
+	for (g = 0; g < USER_PTRS_PER_PGD; g++, pgd++) {
+		if (pgd_none(*pgd))
+			continue;
+		pud = pud_offset(pgd, 0);
+
+		if (PTRS_PER_PUD > 1) /* not folded */
+			pgd_walk_set_prot(pud, flags);
+
+		for (u = 0; u < PTRS_PER_PUD; u++, pud++) {
+			if (pud_none(*pud))
+				continue;
+			pmd = pmd_offset(pud, 0);
+
+			if (PTRS_PER_PMD > 1) /* not folded */
+				pgd_walk_set_prot(pmd, flags);
+
+			for (m = 0; m < PTRS_PER_PMD; m++, pmd++) {
+				if (pmd_none(*pmd))
+					continue;
+
+				/* This can get called before mem_map
+				   is set up, so we assume nothing is
+				   highmem at that point. */
+				if (mem_map == NULL ||
+				    !PageHighMem(pmd_page(*pmd))) {
+					pte = pte_offset_kernel(pmd, 0);
+					pgd_walk_set_prot(pte, flags);
+				}
+			}
+		}
+	}
+
+	if (HYPERVISOR_update_va_mapping((unsigned long)pgd_base,
+					 pfn_pte(PFN_DOWN(__pa(pgd_base)),
+						 flags),
+					 UVMF_TLB_FLUSH) < 0)
+		BUG();
+}
+
+
+/* This is called just after a mm has been duplicated from its parent,
+   but it has not been used yet.  We need to make sure that its
+   pagetable is all read-only, and can be pinned. */
+void xen_pgd_pin(pgd_t *pgd)
+{
+	struct mmuext_op op;
+
+	pgd_walk(pgd, PAGE_KERNEL_RO);
+
+#if defined(CONFIG_X86_PAE)
+	op.cmd = MMUEXT_PIN_L3_TABLE;
+#else
+	op.cmd = MMUEXT_PIN_L2_TABLE;
+#endif
+	op.arg1.mfn = pfn_to_mfn(PFN_DOWN(__pa(pgd)));
+	if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0)
+		BUG();
+}
+
+/* Release a pagetables pages back as normal RW */
+void xen_pgd_unpin(pgd_t *pgd)
+{
+	struct mmuext_op op;
+
+	op.cmd = MMUEXT_UNPIN_TABLE;
+	op.arg1.mfn = pfn_to_mfn(PFN_DOWN(__pa(pgd)));
+
+	if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0)
+		BUG();
+
+	pgd_walk(pgd, PAGE_KERNEL);
+}
+
+
+void xen_activate_mm(struct mm_struct *prev, struct mm_struct *next)
+{
+	xen_pgd_pin(next->pgd);
+}
+
+void xen_dup_mmap(struct mm_struct *oldmm, struct mm_struct *mm)
+{
+	xen_pgd_pin(mm->pgd);
+}
+
+void xen_exit_mmap(struct mm_struct *mm)
+{
+	struct task_struct *tsk = current;
+
+	task_lock(tsk);
+
+	/*
+	 * We aggressively remove defunct pgd from cr3. We execute unmap_vmas()
+	 * *much* faster this way, as no tlb flushes means bigger wrpt batches.
+	 */
+	if (tsk->active_mm == mm) {
+		tsk->active_mm = &init_mm;
+		atomic_inc(&init_mm.mm_count);
+
+		switch_mm(mm, &init_mm, tsk);
+
+		atomic_dec(&mm->mm_count);
+		BUG_ON(atomic_read(&mm->mm_count) == 0);
+	}
+
+	task_unlock(tsk);
+
+	xen_pgd_unpin(mm->pgd);
+}

arch/i386/xen/mmu.h

@@ -0,0 +1,47 @@
+#ifndef _XEN_MMU_H
+
+#include <linux/linkage.h>
+#include <asm/page.h>
+
+void set_pte_mfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags);
+
+void xen_set_pte(pte_t *ptep, pte_t pteval);
+void xen_set_pte_at(struct mm_struct *mm, unsigned long addr,
+		    pte_t *ptep, pte_t pteval);
+void xen_set_pmd(pmd_t *pmdp, pmd_t pmdval);
+
+void xen_activate_mm(struct mm_struct *prev, struct mm_struct *next);
+void xen_dup_mmap(struct mm_struct *oldmm, struct mm_struct *mm);
+void xen_exit_mmap(struct mm_struct *mm);
+
+void xen_pgd_pin(pgd_t *pgd);
+void xen_pgd_unpin(pgd_t *pgd);
+
+#ifdef CONFIG_X86_PAE
+unsigned long long xen_pte_val(pte_t);
+unsigned long long xen_pmd_val(pmd_t);
+unsigned long long xen_pgd_val(pgd_t);
+
+pte_t xen_make_pte(unsigned long long);
+pmd_t xen_make_pmd(unsigned long long);
+pgd_t xen_make_pgd(unsigned long long);
+
+void xen_set_pte_at(struct mm_struct *mm, unsigned long addr,
+		    pte_t *ptep, pte_t pteval);
+void xen_set_pte_atomic(pte_t *ptep, pte_t pte);
+void xen_set_pud(pud_t *ptr, pud_t val);
+void xen_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
+void xen_pmd_clear(pmd_t *pmdp);
+
+
+#else
+unsigned long xen_pte_val(pte_t);
+unsigned long xen_pmd_val(pmd_t);
+unsigned long xen_pgd_val(pgd_t);
+
+pte_t xen_make_pte(unsigned long);
+pmd_t xen_make_pmd(unsigned long);
+pgd_t xen_make_pgd(unsigned long);
+#endif
+
+#endif	/* _XEN_MMU_H */