[Kos-cvs] kos/modules/x86/mm Makefile,1.5,1.6 _vmap.c,1.27,1.28

thomas at kos.enix.org thomas at kos.enix.org
Wed Dec 29 20:28:53 CET 2004

Update of /home/kos/cvs/kos/modules/x86/mm
In directory the-doors:/tmp/cvs-serv1745/modules/x86/mm

Modified Files:
	Makefile _vmap.c 
Log Message:
2004-12-29  Thomas Petazzoni  <thomas at crazy.kos.nx>

	* modules/x86/task/_thread_cpu_context.c
	(init_user_thread_context): VMM functions now work with address

	* modules/x86/mm/_vmap.c (arch_do_unmap_virtual_page): Unmapping a
	page that hasn't any corresponding PT is not an error. So, the
	ASSERT_FATAL() was removed, and replaced by a test that returns
	SUCCESS if there's not PT associated to the page.

	* modules/vmm/vmm.h: VMM functions now operate on address space
	not on teams.

	* modules/vmm/vmm.c: Try to limit the exports only to the modules
	that really need them.

	* modules/vmm/_vmm_map.c: VMM functions now operate on address space,
	not on teams.

	* modules/vmm/_vmm_as.c: VMM functions now operate on address space,
	not on teams. Check the return of kmutex_lock and kmutex_unlock
	for safety.

	* modules/vmm/_vmm.h: VMM functions now operate on address space,
	not on teams.

	* modules/test/vmm_test.c: A couple of updates, whitespaces
	cleanup. In the hope of the stress test to pass, one day, maybe

	* modules/test/test.c: Add the new mutex test.

	* modules/test/sem_test.c: Semaphore test update. Less kernel
	threads are created, less time is spent in usleep() and
	create_kernel_thread() return is checked to make sure all kernel
	threads are correctly created.

	* modules/test/mutex_test.c: New mutex test.

	* modules/test/_test.h: New mutex_test() function.

	* modules/test/Makefile (OBJS): New mutex_test.c file.

	* modules/task/_task_kstack.c (unallocate_cpl0_stack): Update
	calls to unmap_virtual_range() according to new prototype.

	* modules/pci/_pci.c (_pci_scan_bus): Not initializing pca
	(pci_config_address) to 0 is not good ! Now, PCI devices are
	correctly detected under Bochs and Qemu. Even a network adapter,
	along with its I/O space and IRQ configuration can be detected !

	* modules/lib/std/stdlib.h (printk): printk is re-defined as
	ktty_printk, and not bochs_printk. By the way, I find this #define
	quite ugly.

	* modules/kos/wolfgang.c (primary_thread): Testing, testing,
	debugging, debugging, testing...

	* modules/kitc/_kmutex.c: Changed k_ui32_t to
	spinlock_flags_t. Added some debugging messages and assertions.

	* MkVars (KOSSYSPATH): The path to the kos-sys CVS module, in
	which the hard disk image is stored. The default value is
	../kos-sys/, but it can be overriden using the .mkvars file.

	* Makefile (qemu): Instead of an hard-coded path to the hard disk
	image, we use a KOSSYSPATH variable that can be overriden in the
	.mkvars file.

Index: Makefile
RCS file: /home/kos/cvs/kos/modules/x86/mm/Makefile,v
retrieving revision 1.5
retrieving revision 1.6
diff -u -d -r1.5 -r1.6
--- Makefile	28 Dec 2004 18:45:02 -0000	1.5
+++ Makefile	29 Dec 2004 19:28:51 -0000	1.6
@@ -1,5 +1,5 @@
 OBJS= _gdt.o _vmap.o _team_mm_context.o _pgflt.o _rmap.o mm.o
 all: arch_mm.ro

Index: _vmap.c
RCS file: /home/kos/cvs/kos/modules/x86/mm/_vmap.c,v
retrieving revision 1.27
retrieving revision 1.28
diff -u -d -r1.27 -r1.28
--- _vmap.c	28 Dec 2004 18:45:02 -0000	1.27
+++ _vmap.c	29 Dec 2004 19:28:51 -0000	1.28
@@ -162,6 +162,11 @@
+  DEBUG_PRINT1("PD index in PD table : %d => PD 0x%x <=> PD 0x%x ?\n",
+	       dest_mm_ctxt->pd_index_in_pd_table,
+	       PD_ANY_TEAM(dest_mm_ctxt->pd_index_in_pd_table),
+	       dest_mm_ctxt->pd);
   /* Make sure we are mapping the correct foreign PD */
 		     & PAGE_ADDR_MASK) == dest_mm_ctxt->pd);
@@ -593,11 +598,16 @@
-  ASSERT_FATAL(IS_PRESENT(pd[va_to_pde(vaddr)]));
   DEBUG_PRINT3("[arch_do_unmap_virtual_page] Unmapping with pd_index_in_pd_table = %d (0x%x)\n",
 	       pd_index_in_pd_table, dest_mm_ctxt);
+  /* No PT for this page => page is not present nor swapped, so it's
+     already unmapped */
+  if(! IS_PRESENT(pd[va_to_pde(vaddr)]))
+    {
+      return ESUCCESS;
+    }
   /* Page is physically present in RAM (aka, not swapped) */

More information about the Kos-cvs mailing list