Changeset f8ddd17 in mainline for kernel/generic/src/ddi/ddi.c

Timestamp:

2006-12-09T20:20:50Z (18 years ago)

Author:

Jakub Jermar <jakub@…>

Branches:

lfn, master, serial, ticket/834-toolchain-update, topic/msim-upgrade, topic/simplify-dev-export

Children:

b82a13c

Parents:

9ab9c2ec

Message:

Rework support for virtually indexed cache.
Instead of repeatedly flushing the data cache, which was a huge overkill, refuse to create an illegal address alias
in the kernel (again) and allocate appropriate page color in userspace instead. Extend the detection also to
SYS_PHYSMEM_MAP syscall.

Add support for tracking physical memory areas mappable by SYS_PHYSMEM_MAP.

Lots of coding style changes.

File:

• kernel/generic/src/ddi/ddi.c (modified) (5 diffs)

kernel/generic/src/ddi/ddi.c

@@ -48 +48 @@
 #include <synch/spinlock.h>
 #include <syscall/copy.h>
+#include <adt/btree.h>
 #include <arch.h>
 #include <align.h>
 #include <errno.h>
 
+/** This lock protects the parea_btree. */
+SPINLOCK_INITIALIZE(parea_lock);
+
+/** B+tree with enabled physical memory areas. */
+static btree_t parea_btree;
+
+/** Initialize DDI. */
+void ddi_init(void)
+{
+        btree_create(&parea_btree);
+}
+
+/** Enable piece of physical memory for mapping by physmem_map().
+ *
+ * @param parea Pointer to physical area structure.
+ *
+ * @todo This function doesn't check for overlaps. It depends on the kernel to
+ * create disjunct physical memory areas.
+ */
+void ddi_parea_register(parea_t *parea)
+{
+        ipl_t ipl;
+
+        ipl = interrupts_disable();
+        spinlock_lock(&parea_lock);
+        
+        /*
+         * TODO: we should really check for overlaps here.
+         * However, we should be safe because the kernel is pretty sane and
+         * memory of different devices doesn't overlap. 
+         */
+        btree_insert(&parea_btree, (btree_key_t) parea->pbase, parea, NULL);
+
+        spinlock_unlock(&parea_lock);
+        interrupts_restore(ipl);        
+}
+
 /** Map piece of physical memory into virtual address space of current task.
  *
- * @param pf Physical frame address of the starting frame.
- * @param vp Virtual page address of the starting page.
+ * @param pf Physical address of the starting frame.
+ * @param vp Virtual address of the starting page.
  * @param pages Number of pages to map.
  * @param flags Address space area flags for the mapping.
  *
- * @return 0 on success, EPERM if the caller lacks capabilities to use this syscall,
- *         ENOENT if there is no task matching the specified ID and ENOMEM if
- *         there was a problem in creating address space area.
+ * @return 0 on success, EPERM if the caller lacks capabilities to use this
+ *      syscall, ENOENT if there is no task matching the specified ID or the
+ *      physical address space is not enabled for mapping and ENOMEM if there
+ *      was a problem in creating address space area. ENOTSUP is returned when
+ *      an attempt to create an illegal address alias is detected.
  */
 static int ddi_physmem_map(uintptr_t pf, uintptr_t vp, count_t pages, int flags)
@@ -80 +120 @@
 
         ipl = interrupts_disable();
+
+        /*
+         * Check if the physical memory area is enabled for mapping.
+         * If the architecture supports virtually indexed caches, intercept
+         * attempts to create an illegal address alias.
+         */
+        spinlock_lock(&parea_lock);
+        parea_t *parea;
+        btree_node_t *nodep;
+        parea = btree_search(&parea_btree, (btree_key_t) pf, &nodep);
+        if (!parea || parea->frames < pages || ((flags & AS_AREA_CACHEABLE) &&
+                !parea->cacheable) || (!(flags & AS_AREA_CACHEABLE) &&
+                parea->cacheable)) {
+                /*
+                 * This physical memory area cannot be mapped.
+                 */
+                spinlock_unlock(&parea_lock);
+                interrupts_restore(ipl);
+                return ENOENT;
+        }
+
+#ifdef CONFIG_VIRT_IDX_DCACHE
+        if (PAGE_COLOR(parea->vbase) != PAGE_COLOR(vp)) {
+                /*
+                 * Refuse to create an illegal address alias.
+                 */
+                spinlock_unlock(&parea_lock);
+                interrupts_restore(ipl);
+                return ENOTSUP;
+        }
+#endif /* CONFIG_VIRT_IDX_DCACHE */
+
+        spinlock_unlock(&parea_lock);
+
         spinlock_lock(&TASK->lock);
         
@@ -108 +182 @@
  * @param size Size of the enabled I/O space..
  *
- * @return 0 on success, EPERM if the caller lacks capabilities to use this syscall,
- *         ENOENT if there is no task matching the specified ID.
+ * @return 0 on success, EPERM if the caller lacks capabilities to use this
+ *      syscall, ENOENT if there is no task matching the specified ID.
  */
 static int ddi_iospace_enable(task_id_t id, uintptr_t ioaddr, size_t size)
@@ -161 +235 @@
  * @return 0 on success, otherwise it returns error code found in errno.h
  */ 
-unative_t sys_physmem_map(unative_t phys_base, unative_t virt_base, unative_t pages, 
-                         unative_t flags)
-{
-        return (unative_t) ddi_physmem_map(ALIGN_DOWN((uintptr_t) phys_base, FRAME_SIZE),
-                                          ALIGN_DOWN((uintptr_t) virt_base, PAGE_SIZE), (count_t) pages,
-                                          (int) flags);
+unative_t sys_physmem_map(unative_t phys_base, unative_t virt_base, unative_t
+        pages, unative_t flags)
+{
+        return (unative_t) ddi_physmem_map(ALIGN_DOWN((uintptr_t) phys_base,
+                FRAME_SIZE), ALIGN_DOWN((uintptr_t) virt_base, PAGE_SIZE),
+                (count_t) pages, (int) flags);
 }
 
@@ -184 +258 @@
                 return (unative_t) rc;
                 
-        return (unative_t) ddi_iospace_enable((task_id_t) arg.task_id, (uintptr_t) arg.ioaddr, (size_t) arg.size);
+        return (unative_t) ddi_iospace_enable((task_id_t) arg.task_id,
+                (uintptr_t) arg.ioaddr, (size_t) arg.size);
 }
 
 /** Disable or enable preemption.
  *
- * @param enable If non-zero, the preemption counter will be decremented, leading to potential
- *               enabling of preemption. Otherwise the preemption counter will be incremented,
- *               preventing preemption from occurring.
+ * @param enable If non-zero, the preemption counter will be decremented,
+ *      leading to potential enabling of preemption. Otherwise the preemption
+ *      counter will be incremented, preventing preemption from occurring.
  *
  * @return Zero on success or EPERM if callers capabilities are not sufficient.