kernel: vm: Add initial demand paging support

Signed-off-by: Ian Moffett <ian@osmora.org>

7 files changed, 250 insertions, 18 deletions

diff --git a/sys/arch/amd64/amd64/trap.S b/sys/arch/amd64/amd64/trap.S
index 66dd2a9..3cf6888 100644
--- a/sys/arch/amd64/amd64/trap.S
+++ b/sys/arch/amd64/amd64/trap.S
@@ -131,9 +131,8 @@ page_fault:
 
     handle_trap
 
-    /* TODO */
-    cli
-    hlt
+    pop_trapframe
+    iretq
 
 .globl nmi
 nmi:
diff --git a/sys/arch/amd64/amd64/trap.c b/sys/arch/amd64/amd64/trap.c
index 3caaaff..be0493f 100644
--- a/sys/arch/amd64/amd64/trap.c
+++ b/sys/arch/amd64/amd64/trap.c
@@ -35,6 +35,7 @@
 #include <sys/signal.h>
 #include <sys/proc.h>
 #include <sys/sched.h>
+#include <vm/fault.h>
 
 static const char *trap_type[] = {
     [TRAP_BREAKPOINT]   = "breakpoint",
@@ -53,6 +54,31 @@ static const char *trap_type[] = {
 
 static const int TRAP_COUNT = __ARRAY_COUNT(trap_type);
 
+
+static inline vaddr_t
+pf_faultaddr(void)
+{
+    uintptr_t cr2;
+    __ASMV("mov %%cr2, %0\n" : "=r" (cr2) :: "memory");
+    return cr2;
+}
+
+static inline vm_prot_t
+pf_accesstype(struct trapframe *tf)
+{
+    vm_prot_t prot = 0;
+    uint64_t ec = tf->error_code;
+
+    if (__TEST(ec, __BIT(1)))
+        prot |= PROT_WRITE;
+    if (__TEST(ec, __BIT(2)))
+        prot |= PROT_USER;
+    if (__TEST(ec, __BIT(4)))
+        prot |= PROT_EXEC;
+
+    return prot;
+}
+
 static void
 dbg_errcode(struct trapframe *tf)
 {
@@ -87,11 +113,10 @@ trap_print(struct trapframe *tf)
 static void
 regdump(struct trapframe *tf)
 {
-    uintptr_t cr3, cr2;
+    uintptr_t cr3, cr2 = pf_faultaddr();
 
-    __ASMV("mov %%cr2, %0\n"
-           "mov %%cr3, %1\n"
-           : "=r" (cr2), "=r" (cr3)
+    __ASMV("mov %%cr3, %0\n"
+           : "=r" (cr3)
            :
            : "memory"
     );
@@ -121,6 +146,7 @@ void
 trap_handler(struct trapframe *tf)
 {
     struct proc *curtd = this_td();
+    int s;
 
     /*
      * XXX: Handle NMIs better. For now we just
@@ -142,6 +168,11 @@ trap_handler(struct trapframe *tf)
     case TRAP_ARITH_ERR:
         signal_raise(curtd, SIGFPE);
         break;
+    case TRAP_PAGEFLT:
+        s = vm_fault(pf_faultaddr(), pf_accesstype(tf));
+        if (s != 0)
+            signal_raise(curtd, SIGSEGV);
+        break;
     default:
         signal_raise(curtd, SIGSEGV);
         break;
diff --git a/sys/include/vm/fault.h b/sys/include/vm/fault.h
new file mode 100644
index 0000000..f0e308c
--- /dev/null
+++ b/sys/include/vm/fault.h
@@ -0,0 +1,38 @@
+/*
+ * Copyright (c) 2023-2024 Ian Marco Moffett and the Osmora Team.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the name of Hyra nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _VM_FAULT_H_
+#define _VM_FAULT_H_
+
+#include <sys/types.h>
+#include <vm/pmap.h>
+
+int vm_fault(vaddr_t va, vm_prot_t access_type);
+
+#endif  /* !_VM_FAULT_H_ */
diff --git a/sys/include/vm/map.h b/sys/include/vm/map.h
index f482788..1da8240 100644
--- a/sys/include/vm/map.h
+++ b/sys/include/vm/map.h
@@ -46,10 +46,14 @@
 /* Memory map table entry count */
 #define MTAB_ENTRIES 32
 
+struct vm_object;
+
 struct vm_mapping {
     TAILQ_ENTRY(vm_mapping) link;
     struct vm_range range;
+    struct vm_object *vmobj;
     paddr_t physmem_base;
+    vm_prot_t prot;
 
     /* Private */
     size_t vhash;           /* Virtual address hash */
diff --git a/sys/include/vm/obj.h b/sys/include/vm/obj.h
index c1c2f17..ef07a1e 100644
--- a/sys/include/vm/obj.h
+++ b/sys/include/vm/obj.h
@@ -42,6 +42,7 @@ struct vm_object {
     struct vm_pagerops *pgops;      /* Pager operations */
 
     uint8_t is_anon : 1;            /* Is an anonymous mapping */
+    uint8_t demand  : 1;            /* Only mapped upon access */
     int ref;                        /* Ref count */
     struct vnode *vnode;            /* Only used if `is_anon` is 0 */
 };
diff --git a/sys/vm/vm_fault.c b/sys/vm/vm_fault.c
new file mode 100644
index 0000000..ec733e8
--- /dev/null
+++ b/sys/vm/vm_fault.c
@@ -0,0 +1,114 @@
+/*
+ * Copyright (c) 2023-2024 Ian Marco Moffett and the Osmora Team.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the name of Hyra nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <sys/types.h>
+#include <sys/sched.h>
+#include <vm/fault.h>
+#include <vm/map.h>
+#include <vm/pmap.h>
+#include <vm/vm.h>
+#include <vm/physseg.h>
+
+static struct vm_mapping *
+vm_mapq_search(vm_mapq_t *mq, vaddr_t addr)
+{
+    struct vm_mapping *mapping;
+    const struct vm_range *range;
+
+    TAILQ_FOREACH(mapping, mq, link) {
+        range = &mapping->range;
+        if (addr >= range->start && addr <= range->end) {
+            return mapping;
+        }
+    }
+
+    return NULL;
+}
+
+static struct vm_mapping *
+vm_find_mapping(vaddr_t addr)
+{
+    struct vm_mapping *mapping;
+    struct proc *td;
+    vm_mapq_t *mapq;
+
+    mapping = vm_mapping_fetch(&td->mapspace, addr);
+    if (mapping != NULL)
+        return mapping;
+
+    /* Need to search other maps */
+    td = this_td();
+    for (size_t i = 0; i < MTAB_ENTRIES; ++i) {
+        mapq = &td->mapspace.mtab[i];
+        mapping = vm_mapq_search(mapq, addr);
+        if (mapping != NULL)
+            return mapping;
+    }
+
+    return NULL;
+}
+
+int
+vm_fault(vaddr_t va, vm_prot_t access_type)
+{
+    struct proc *td = this_td();
+    struct vm_mapping *mapping;
+    struct vm_object *vmobj;
+
+    size_t granule = vm_get_page_size();
+    vaddr_t va_base = va &= ~(granule - 1);
+
+    int s;
+    paddr_t pa_base;
+
+    mapping = vm_find_mapping(va_base);
+    if (mapping == NULL)
+        return -1;
+
+    if ((vmobj = mapping->vmobj) == NULL)
+        /* Virtual memory object non-existent */
+        return -1;
+    if (!vmobj->demand)
+        /* Demand paging not enabled for this object */
+        return -1;
+    if ((access_type & ~mapping->prot) != 0)
+        /* Invalid access type */
+        return -1;
+
+    /* Allocate physical memory if needed */
+    if (mapping->physmem_base == 0) {
+        pa_base = vm_alloc_pageframe(1);
+        mapping->physmem_base = pa_base;
+    } else {
+        pa_base = mapping->physmem_base;
+    }
+
+    s = vm_map_create(td->addrsp, va_base, pa_base, access_type, granule);
+    return s;
+}
diff --git a/sys/vm/vm_map.c b/sys/vm/vm_map.c
index 8f7accc..7e67ebd 100644
--- a/sys/vm/vm_map.c
+++ b/sys/vm/vm_map.c
@@ -140,7 +140,7 @@ vm_mapspace_insert(struct vm_mapspace *ms, struct vm_mapping *mapping)
 }
 
 /*
- * Create an anonymous mapping.
+ * Create a mapping (internal helper)
  *
  * @addr: Address to map.
  * @physmem: Physical address, set to 0 to alloc one here
@@ -149,7 +149,7 @@ vm_mapspace_insert(struct vm_mapspace *ms, struct vm_mapping *mapping)
  * Returns zero on failure.
  */
 static paddr_t
-vm_anon_map(void *addr, paddr_t physmem, vm_prot_t prot, size_t len)
+vm_map(void *addr, paddr_t physmem, vm_prot_t prot, size_t len)
 {
     struct proc *td = this_td();
     const size_t GRANULE = vm_get_page_size();
@@ -187,7 +187,8 @@ vm_anon_map(void *addr, paddr_t physmem, vm_prot_t prot, size_t len)
  * @fd: File descriptor.
  */
 static paddr_t
-vm_fd_map(void *addr, vm_prot_t prot, size_t len, off_t off, int fd)
+vm_fd_map(void *addr, vm_prot_t prot, size_t len, off_t off, int fd,
+          struct vm_mapping *mapping)
 {
     paddr_t physmem = 0;
 
@@ -216,6 +217,9 @@ vm_fd_map(void *addr, vm_prot_t prot, size_t len, off_t off, int fd)
     if (vm_obj_init(&vp->vmobj, vp) != 0)
         return 0;
 
+    mapping->vmobj = vp->vmobj;
+    vm_object_ref(vp->vmobj);
+
     /* Try to fetch a physical address */
     if (vm_pager_paddr(vp->vmobj, &physmem, prot) != 0) {
         vm_obj_destroy(vp->vmobj);
@@ -228,7 +232,7 @@ vm_fd_map(void *addr, vm_prot_t prot, size_t len, off_t off, int fd)
      * then connect it to the physical address (creates a shared mapping)
      */
     if (physmem != 0) {
-        vm_anon_map(addr, physmem, prot, len);
+        vm_map(addr, physmem, prot, len);
         return physmem;
     }
 
@@ -238,7 +242,7 @@ vm_fd_map(void *addr, vm_prot_t prot, size_t len, off_t off, int fd)
      * anonymous mapping then page-in from whatever filesystem
      * (creates a shared mapping)
      */
-    physmem = vm_anon_map(addr, 0, prot, len);
+    physmem = vm_map(addr, 0, prot, len);
     pg.physaddr = physmem;
 
     if (vm_pager_get(vp->vmobj, off, len, &pg) != 0) {
@@ -255,6 +259,9 @@ munmap(void *addr, size_t len)
     struct proc *td = this_td();
     struct vm_mapping *mapping;
 
+    struct vm_object *obj;
+    struct vnode *vp;
+
     struct vm_mapspace *ms;
     size_t map_len, granule;
     vaddr_t map_start, map_end;
@@ -272,6 +279,23 @@ munmap(void *addr, size_t len)
     map_end = mapping->range.end;
     map_len = map_end - map_start;
 
+    /* Try to release any virtual memory objects */
+    if ((obj = mapping->vmobj) != NULL) {
+        spinlock_acquire(&obj->lock);
+        /*
+         * Drop our ref and try to cleanup. If the refcount
+         * is > 1, something is still holding it and we can't
+         * do much.
+         */
+        vm_object_unref(obj);
+        vp = obj->vnode;
+        if (vp != NULL && obj->ref == 1) {
+            vp->vmobj = NULL;
+            vm_obj_destroy(obj);
+        }
+        spinlock_release(&obj->lock);
+    }
+
     /* Release the mapping */
     vm_map_destroy(td->addrsp, map_start, map_len);
     vm_free_pageframe(mapping->range.start, map_len / granule);
@@ -292,10 +316,13 @@ mmap(void *addr, size_t len, int prot, int flags, int fildes, off_t off)
 
     struct proc *td = this_td();
     struct vm_mapping *mapping = ALLOC_MAPPING();
+    struct vm_object *vmobj;
 
     size_t misalign = ((vaddr_t)addr) & (GRANULE - 1);
     paddr_t physmem = 0;
 
+    mapping->prot = prot | PROT_USER;
+
     /* Ensure of valid prot flags */
     if ((prot & ~PROT_MASK) != 0)
         return MAP_FAILED;
@@ -315,15 +342,33 @@ mmap(void *addr, size_t len, int prot, int flags, int fildes, off_t off)
      * this is.
      */
     if (__TEST(flags, MAP_ANONYMOUS)) {
-        /* Handle an anonymous map request */
-        physmem = vm_anon_map(addr, 0, prot, len);
+        /* Try to create a virtual memory object */
+        if (vm_obj_init(&vmobj, NULL) != 0)
+            return 0;
+
+        /*
+         * Enable demand paging for this object if
+         * `addr` is not NULL.
+         */
+        if (addr != NULL) {
+            vmobj->is_anon = 1;
+            vmobj->demand = 1;
+
+            mapping->vmobj = vmobj;
+            mapping->physmem_base = 0;
+        } else {
+            physmem = vm_map(addr, 0, prot, len);
+        }
+
+        /* Did this work? */
+        if (physmem == 0 && addr == NULL)
+            return MAP_FAILED;
     } else if (__TEST(flags, MAP_SHARED)) {
-        physmem = vm_fd_map(addr, prot, len, off, fildes);
+        physmem = vm_fd_map(addr, prot, len, off, fildes, mapping);
+        if (physmem == 0)
+            return MAP_FAILED;
     }
 
-    if (physmem == 0) {
-        return MAP_FAILED;
-    }
 
     map_start = __ALIGN_DOWN((vaddr_t)addr, GRANULE);
     map_end = map_start + __ALIGN_UP(len + misalign, GRANULE);