don't always copy for i/o instructions.

processes than the caller..

also disable kernel sanity checks

drivers/at_wini/at_wini.c

@@ -493,7 +493,8 @@ PRIVATE void init_params()
 	dma_buf = mmap(0, ATA_DMA_BUF_SIZE, PROT_READ|PROT_WRITE,
 		MAP_PREALLOC | MAP_CONTIG | MAP_ANON, -1, 0);
 	prdt = mmap(0, PRDT_BYTES,
-		PROT_READ|PROT_WRITE, MAP_CONTIG | MAP_ANON, -1, 0);
+		PROT_READ|PROT_WRITE,
+		MAP_PREALLOC | MAP_CONTIG | MAP_ANON, -1, 0);
 	if(dma_buf == MAP_FAILED || prdt == MAP_FAILED) {
 		disable_dma = 1;
 		printf("at_wini%d: no dma\n", w_instance);

drivers/lance/lance.c

@@ -522,6 +522,7 @@ pci_init();
 	  }
 	  else
 	  {
+		printf("lance buf: 0x%lx\n", vir2phys(lance_buf));
 	  	report( "LANCE", "DMA denied because address out of range", NO_NUM );
 	  }
 	  

drivers/memory/memory.c

@@ -403,7 +403,8 @@ int safe;
 #endif
 
 	/* Try to allocate a piece of memory for the RAM disk. */
-	if((mem = mmap(0, ramdev_size, PROT_READ|PROT_WRITE, MAP_ANON, -1, 0)) == MAP_FAILED) {
+	if((mem = mmap(0, ramdev_size, PROT_READ|PROT_WRITE,
+		MAP_PREALLOC|MAP_ANON, -1, 0)) == MAP_FAILED) {
 	    printf("MEM: failed to get memory for ramdisk\n");
             return(ENOMEM);
         } 

drivers/pci/pci.c

@@ -33,8 +33,8 @@ Created:	Jan 2000 by Philip Homburg <philip@cs.vu.nl>
 #include <string.h>
 #include <minix/sysutil.h>
 
-#define NR_PCIBUS	10
-#define NR_PCIDEV	40
+#define NR_PCIBUS       40
+#define NR_PCIDEV       50
 
 #define PBT_INTEL_HOST	 1
 #define PBT_PCIBRIDGE	 2
@@ -1135,16 +1135,24 @@ int bar_nr;
 {
 	int reg, prefetch, type, dev_bar_nr;
 	u32_t bar, bar2;
+	u16_t cmd;
 
 	reg= PCI_BAR+4*bar_nr;
 
 	bar= pci_attr_r32_u(devind, reg);
 	if (bar & PCI_BAR_IO)
 	{
-		/* Size register */
+		/* Disable I/O access before probing for BAR's size */
+		cmd = pci_attr_r16(devind, PCI_CR);
+		pci_attr_w16(devind, PCI_CR, cmd & ~PCI_CR_IO_EN);
+
+		/* Probe BAR's size */
 		pci_attr_w32(devind, reg, 0xffffffff);
 		bar2= pci_attr_r32_u(devind, reg);
+
+		/* Restore original state */
 		pci_attr_w32(devind, reg, bar);
+		pci_attr_w16(devind, PCI_CR, cmd);
 
 		bar &= ~(u32_t)3;	/* Clear non-address bits */
 		bar2 &= ~(u32_t)3;
@@ -1168,10 +1176,17 @@ int bar_nr;
 	}
 	else
 	{
-		/* Size register */
+		/* Disable mem access before probing for BAR's size */
+		cmd = pci_attr_r16(devind, PCI_CR);
+		pci_attr_w16(devind, PCI_CR, cmd & ~PCI_CR_MEM_EN);
+
+		/* Probe BAR's size */
 		pci_attr_w32(devind, reg, 0xffffffff);
 		bar2= pci_attr_r32_u(devind, reg);
+
+		/* Restore original values */
 		pci_attr_w32(devind, reg, bar);
+		pci_attr_w16(devind, PCI_CR, cmd);
 
 		if (bar2 == 0)
 			return;	/* Reg. is not implemented */
@@ -2398,7 +2413,7 @@ u32_t value;
 #if 0
 	printf("pcii_wreg32(%d, %d, 0x%X, 0x%X): %d.%d.%d\n",
 		busind, devind, port, value,
-		pcibus[busind].pb_bus, pcidev[devind].pd_dev,
+		pcibus[busind].pb_busnr, pcidev[devind].pd_dev,
 		pcidev[devind].pd_func);
 #endif
 	PCII_WREG32_(pcibus[busind].pb_busnr, 

drivers/random/main.c

@@ -305,7 +305,7 @@ message *m_ptr;				/* pointer to alarm message */
   random_update(RND_TIMING, &r, 1);
 
   /* Schedule new alarm for next m_random call. */
-  if (OK != (s=sys_setalarm(KRANDOM_PERIOD, 0)))
+  if (OK != (s=sys_setalarm(random_isseeded() ? KRANDOM_PERIOD : sys_hz(), 0)))
   	report("RANDOM", "sys_setalarm failed", s);
 }
 

drivers/tty/console.c

@@ -52,7 +52,6 @@ PRIVATE unsigned font_lines;	/* font lines per character */
 PRIVATE unsigned scr_width;	/* # characters on a line */
 PRIVATE unsigned scr_lines;	/* # lines on the screen */
 PRIVATE unsigned scr_size;	/* # characters on the screen */
-PUBLIC unsigned info_location;	/* location in video memory of struct */
 
 /* tells mem_vid_copy() to blank the screen */
 #define BLANK_MEM ((vir_bytes) 0) 
@@ -66,9 +65,6 @@ PRIVATE int disabled_sm;	/* Scroll mode to be restored when re-enabling
 
 char *console_memory = NULL;
 
-/* boot_tty_info we use to communicate with the boot code. */
-struct boot_tty_info boot_tty_info;
-
 /* Per console data. */
 typedef struct console {
   tty_t *c_tty;			/* associated TTY struct */
@@ -90,24 +86,14 @@ typedef struct console {
   int c_line;			/* line no */
 } console_t;
 
-#define UPDATEBOOTINFO(ccons, infofield, value)	{		\
-	if(ccons->c_line == 0) {				\
-		boot_tty_info.infofield = value;		\
-		mem_vid_copy((vir_bytes) &boot_tty_info,	\
-			info_location/2, sizeof(boot_tty_info)/2);	\
-	}							\
-}
-
 #define UPDATE_CURSOR(ccons, cursor) {				\
 	ccons->c_cur = cursor;					\
-	UPDATEBOOTINFO(ccons, conscursor, ccons->c_cur);	\
 	if(curcons && ccons == curcons)				\
 		set_6845(CURSOR, ccons->c_cur);			\
 }
 
 #define UPDATE_ORIGIN(ccons, origin) {				\
 	ccons->c_org = origin;					\
-	UPDATEBOOTINFO(ccons, consorigin, ccons->c_org);	\
   	if (curcons && ccons == curcons) 			\
 		set_6845(VID_ORG, ccons->c_org);		\
 }
@@ -1001,13 +987,15 @@ tty_t *tp;
   	}
   	if (machine.vdu_ega) vid_size = EGA_SIZE;
   	wrap = ! machine.vdu_ega;
-	info_location = vid_size - sizeof(struct boot_tty_info);
 
 	console_memory = vm_map_phys(SELF, (void *) vid_base, vid_size);
 
 	if(console_memory == MAP_FAILED) 
   		panic("TTY","Console couldn't map video memory", NO_NUM);
 
+	printf("TTY: vm_map_phys of 0x%lx OK, result 0x%lx", 
+		vid_base, console_memory);
+
   	vid_size >>= 1;		/* word count */
   	vid_mask = vid_size - 1;
 
@@ -1015,7 +1003,7 @@ tty_t *tp;
   	scr_size = scr_lines * scr_width;
 
   	/* There can be as many consoles as video memory allows. */
-  	nr_cons = (vid_size - sizeof(struct boot_tty_info)/2) / scr_size;
+  	nr_cons = vid_size / scr_size;
 
   	if (nr_cons > NR_CONS) nr_cons = NR_CONS;
   	if (nr_cons > 1) wrap = 0;
@@ -1038,12 +1026,6 @@ tty_t *tp;
 	scroll_screen(cons, SCROLL_UP);
 	cons->c_row = scr_lines - 1;
 	cons->c_column = 0;
-
-	memset(&boot_tty_info, 0, sizeof(boot_tty_info));
-	UPDATE_CURSOR(cons, cons->c_cur);
-	boot_tty_info.flags = BTIF_CONSCURSOR | BTIF_CONSORIGIN;
-	boot_tty_info.magic = TTYMAGIC;
-	UPDATE_ORIGIN(cons, cons->c_org);
   }
   select_console(0);
   cons_ioctl(tp, 0);

drivers/tty/keyboard.c

@@ -766,8 +766,22 @@ int scode;			/* scan code of key just struck or released */
 			return -1;
 		if(ch)
 			return ch;
-  		printf("tty: ignoring unrecognized %s scancode 0x%x\n",
+		{
+			static char seen[2][NR_SCAN_CODES];
+			int notseen = 0, ei;
+			ei = escape ? 1 : 0;
+			if(scode >= 0 && scode < NR_SCAN_CODES) {
+				notseen = !seen[ei][scode];
+				seen[ei][scode] = 1;
+			} else {
+				printf("tty: scode %d makes no sense\n", scode);
+			}
+			if(notseen) {
+		  		printf("tty: ignoring unrecognized %s "
+					"scancode 0x%x\n",
   				escape ? "escaped" : "straight", scode);
+			}
+		}
   		return -1;
   }
 

include/ibm/pci.h

@@ -9,6 +9,7 @@ Created:	Jan 2000 by Philip Homburg <philip@cs.vu.nl>
 #define PCI_DID		0x02	/* Device ID, 16-bit */
 #define PCI_CR		0x04	/* Command Register, 16-bit */
 #define		PCI_CR_MAST_EN	0x0004	/* Enable Busmaster Access */
+#define		PCI_CR_MEM_EN	0x0002	/* Enable Mem Cycles */
 #define		PCI_CR_IO_EN	0x0001	/* Enable I/O Cycles */
 #define PCI_SR		0x06	/* PCI status, 16-bit */
 #define		 PSR_SSE	0x4000	/* Signaled System Error */

include/minix/com.h

@@ -311,7 +311,6 @@
 #  define SYS_GETINFO    (KERNEL_CALL + 26) 	/* sys_getinfo() */
 #  define SYS_ABORT      (KERNEL_CALL + 27)	/* sys_abort() */
 #  define SYS_IOPENABLE  (KERNEL_CALL + 28)	/* sys_enable_iop() */
-#  define SYS_VM_SETBUF  (KERNEL_CALL + 29)	/* sys_vm_setbuf() */
 #  define SYS_SAFECOPYFROM (KERNEL_CALL + 31)	/* sys_safecopyfrom() */
 #  define SYS_SAFECOPYTO   (KERNEL_CALL + 32)	/* sys_safecopyto() */
 #  define SYS_VSAFECOPY  (KERNEL_CALL + 33)	/* sys_vsafecopy() */
@@ -526,6 +525,7 @@
 				 * and sys_fork
 				 */
 #define PR_FORK_FLAGS	m1_i3
+#define PR_FORK_MSGADDR m1_p1
 
 /* Field names for SYS_INT86 */
 #define INT86_REG86    m1_p1	/* pointer to registers */
@@ -585,6 +585,7 @@
 #define SVMCTL_MRG_LEN		m1_i1	/* MEMREQ_GET reply: length */
 #define SVMCTL_MRG_WRITE	m1_i2	/* MEMREQ_GET reply: writeflag */
 #define SVMCTL_MRG_EP		m1_i3	/* MEMREQ_GET reply: process */
+#define SVMCTL_MRG_REQUESTOR	m1_p2	/* MEMREQ_GET reply: requestor */
 
 /* Codes and field names for SYS_SYSCTL. */
 #define SYSCTL_CODE		m1_i1	/* SYSCTL_CODE_* below */
@@ -603,6 +604,11 @@
 #define VMCTL_MEMREQ_REPLY	15
 #define VMCTL_INCSP		16
 #define VMCTL_NOPAGEZERO	18
+#define VMCTL_I386_KERNELLIMIT	19
+#define VMCTL_I386_PAGEDIRS	20
+#define VMCTL_I386_FREEPDE	23
+#define VMCTL_ENABLE_PAGING	24
+#define VMCTL_I386_INVLPG	25
 
 /*===========================================================================*
  *                Messages for the Reincarnation Server 		     *

include/minix/config.h

@@ -95,7 +95,4 @@
 #define SPROFILE          1    /* statistical profiling */
 #define CPROFILE          0    /* call profiling */
 
-/* Compile kernel so that first page of code and data can be unmapped. */
-#define VM_KERN_NOPAGEZERO      1       
-
 #endif /* _CONFIG_H */

include/minix/debug.h

@@ -0,0 +1,13 @@
+#ifndef _MINIX_DEBUG_H
+#define _MINIX_DEBUG_H 1
+
+/* For reminders of things to be fixed. */
+#define FIXME(str) { static int fixme_warned = 0; \
+	if(!fixme_warned) { \
+		printf("FIXME: %s:%d: %s\n", __FILE__, __LINE__, str);\
+		fixme_warned = 1; \
+	} \
+}
+
+#endif /* _MINIX_DEBUG_H */
+

include/minix/syslib.h

@@ -35,7 +35,7 @@ _PROTOTYPE( int sys_enable_iop, (endpoint_t proc));
 _PROTOTYPE( int sys_exec, (endpoint_t proc, char *ptr,  
 				char *aout, vir_bytes initpc));
 _PROTOTYPE( int sys_fork, (endpoint_t parent, endpoint_t child, int *,
-	struct mem_map *ptr, u32_t vm));
+	struct mem_map *ptr, u32_t vm, vir_bytes *));
 _PROTOTYPE( int sys_newmap, (endpoint_t proc, struct mem_map *ptr));
 _PROTOTYPE( int sys_exit, (endpoint_t proc));
 _PROTOTYPE( int sys_trace, (int req, endpoint_t proc, long addr, long *data_p));
@@ -53,9 +53,8 @@ _PROTOTYPE( int sys_vmctl, (endpoint_t who, int param, u32_t value));
 _PROTOTYPE( int sys_vmctl_get_pagefault_i386, (endpoint_t *who, u32_t *cr2, u32_t *err));
 _PROTOTYPE( int sys_vmctl_get_cr3_i386, (endpoint_t who, u32_t *cr3)  );
 _PROTOTYPE( int sys_vmctl_get_memreq, (endpoint_t *who, vir_bytes *mem,
-        vir_bytes *len, int *wrflag) );
-
-
+        vir_bytes *len, int *wrflag, endpoint_t *) );
+_PROTOTYPE( int sys_vmctl_enable_paging, (struct mem_map *));
 
 _PROTOTYPE( int sys_readbios, (phys_bytes address, void *buf, size_t size));
 _PROTOTYPE( int sys_stime, (time_t boottime));

include/minix/type.h

@@ -97,9 +97,6 @@ struct kinfo {
   int nr_tasks;			/* number of kernel tasks */
   char release[6];		/* kernel release number */
   char version[6];		/* kernel version number */
-#if DEBUG_LOCK_CHECK
-  int relocking;		/* interrupt locking depth (should be 0) */
-#endif
 };
 
 /* Load data accounted every this no. of seconds. */

include/sys/vm_i386.h

@@ -14,6 +14,7 @@ sys/vm_i386.h
 #define I386_VM_ACC	0x020	/* Accessed */
 #define I386_VM_ADDR_MASK 0xFFFFF000 /* physical address */
 #define I386_VM_ADDR_MASK_4MB 0xFFC00000 /* physical address */
+#define I386_VM_OFFSET_MASK_4MB 0x003FFFFF /* physical address */
 
 /* Page directory specific flags. */
 #define I386_VM_BIGPAGE	0x080	/* 4MB page */
@@ -35,6 +36,12 @@ sys/vm_i386.h
 #define I386_VM_PFA_SHIFT	22	/* Page frame address shift */
 
 /* CR0 bits */
+#define I386_CR0_PE		0x00000001	/* Protected mode  */
+#define I386_CR0_MP		0x00000002	/* Monitor Coprocessor  */
+#define I386_CR0_EM		0x00000004	/* Emulate  */
+#define I386_CR0_TS		0x00000008	/* Task Switched  */
+#define I386_CR0_ET		0x00000010	/* Extension Type  */
+#define I386_CR0_WP		0x00010000	/* Enable paging */
 #define I386_CR0_PG		0x80000000	/* Enable paging */
 
 /* some CR4 bits */

kernel/arch/i386/arch_do_vmctl.c

@@ -10,7 +10,9 @@
 #include "../../system.h"
 #include <minix/type.h>
 
-extern u32_t kernel_cr3;
+#include "proto.h"
+
+extern u32_t *vm_pagedirs;
 
 /*===========================================================================*
  *				arch_do_vmctl				     *
@@ -30,7 +32,7 @@ struct proc *p;
 			p->p_seg.p_cr3 = m_ptr->SVMCTL_VALUE;
 			p->p_misc_flags |= MF_FULLVM;
 		} else {
-			p->p_seg.p_cr3 = kernel_cr3;
+			p->p_seg.p_cr3 = 0;
 			p->p_misc_flags &= ~MF_FULLVM;
 		}
 		RTS_LOCK_UNSET(p, VMINHIBIT);
@@ -53,7 +55,27 @@ struct proc *p;
 		m_ptr->SVMCTL_PF_I386_ERR = rp->p_pagefault.pf_flags;
 		return OK;
 	}
+	case VMCTL_I386_KERNELLIMIT:
+	{
+		int r;
+		/* VM wants kernel to increase its segment. */
+		r = prot_set_kern_seg_limit(m_ptr->SVMCTL_VALUE);
+		return r;
 	}
+	case VMCTL_I386_PAGEDIRS:
+	{
+		int pde;
+		vm_pagedirs = (u32_t *) m_ptr->SVMCTL_VALUE;
+		return OK;
+	}
+	case VMCTL_I386_FREEPDE:
+	{
+		i386_freepde(m_ptr->SVMCTL_VALUE);
+		return OK;
+	}
+  }
+
+
 
   kprintf("arch_do_vmctl: strange param %d\n", m_ptr->SVMCTL_PARAM);
   return EINVAL;

kernel/arch/i386/do_sdevio.c

@@ -24,6 +24,8 @@
 PUBLIC int do_sdevio(m_ptr)
 register message *m_ptr;	/* pointer to request message */
 {
+  vir_bytes newoffset;
+  endpoint_t newep;
   int proc_nr, proc_nr_e = m_ptr->DIO_VEC_ENDPT;
   int count = m_ptr->DIO_VEC_SIZE;
   long port = m_ptr->DIO_PORT;
@@ -32,6 +34,9 @@ register message *m_ptr;	/* pointer to request message */
   struct proc *rp;
   struct priv *privp;
   struct io_range *iorp;
+  int rem;
+  vir_bytes addr;
+  struct proc *destproc;
 
   /* Allow safe copies and accesses to SELF */
   if ((m_ptr->DIO_REQUEST & _DIO_SAFEMASK) != _DIO_SAFE &&
@@ -64,11 +69,23 @@ register message *m_ptr;	/* pointer to request message */
   /* Check for 'safe' variants. */
   if((m_ptr->DIO_REQUEST & _DIO_SAFEMASK) == _DIO_SAFE) {
      /* Map grant address to physical address. */
-     if ((phys_buf = umap_verify_grant(proc_addr(proc_nr), who_e,
+     if(verify_grant(proc_nr_e, who_e, 
 	(vir_bytes) m_ptr->DIO_VEC_ADDR,
-	(vir_bytes) m_ptr->DIO_OFFSET, count,
-	req_dir == _DIO_INPUT ? CPF_WRITE : CPF_READ)) == 0)
-         return(EPERM);
+	count,
+	req_dir == _DIO_INPUT ? CPF_WRITE : CPF_READ,
+	(vir_bytes) m_ptr->DIO_OFFSET, 
+	&newoffset, &newep) != OK) {
+	printf("do_sdevio: verify_grant failed\n");
+	return EPERM;
+    }
+	if(!isokendpt(newep, &proc_nr))
+		return(EINVAL);
+     destproc = proc_addr(proc_nr);
+     if ((phys_buf = umap_local(destproc, D,
+	 (vir_bytes) newoffset, count)) == 0) {
+	printf("do_sdevio: umap_local failed\n");
+         return(EFAULT);
+     }
   } else {
      if(proc_nr != who_p)
      {
@@ -77,10 +94,14 @@ register message *m_ptr;	/* pointer to request message */
 	return EPERM;
      }
      /* Get and check physical address. */
-     if ((phys_buf = umap_virtual(proc_addr(proc_nr), D,
+     if ((phys_buf = umap_local(proc_addr(proc_nr), D,
 	 (vir_bytes) m_ptr->DIO_VEC_ADDR, count)) == 0)
          return(EFAULT);
+     destproc = proc_addr(proc_nr);
   }
+     /* current process must be target for phys_* to be OK */
+
+  vm_set_cr3(destproc);
 
 	switch (io_type)
 	{

kernel/arch/i386/exception.c

@@ -10,72 +10,130 @@
 #include <string.h>
 #include <minix/sysutil.h>
 #include "../../proc.h"
+#include "../../proto.h"
+#include "../../vm.h"
 
-extern int vm_copy_in_progress;
+extern int vm_copy_in_progress, catch_pagefaults;
 extern struct proc *vm_copy_from, *vm_copy_to;
-extern u32_t vm_copy_from_v, vm_copy_to_v;
-extern u32_t vm_copy_from_p, vm_copy_to_p, vm_copy_cr3;
 
-u32_t pagefault_cr2, pagefault_count = 0;
-
-void pagefault(struct proc *pr, int trap_errno)
+void pagefault(vir_bytes old_eip, struct proc *pr, int trap_errno,
+	u32_t *old_eipptr, u32_t *old_eaxptr, u32_t pagefaultcr2)
 {
 	int s;
 	vir_bytes ph;
 	u32_t pte;
+	int procok = 0, pcok = 0, rangeok = 0;
+	int in_physcopy = 0;
+	vir_bytes test_eip;
 
-	if(pagefault_count != 1)
-		minix_panic("recursive pagefault", pagefault_count);
+	vmassert(old_eipptr);
+	vmassert(old_eaxptr);
 
-	/* Don't schedule this process until pagefault is handled. */
-	if(RTS_ISSET(pr, PAGEFAULT))
-		minix_panic("PAGEFAULT set", pr->p_endpoint);
-	RTS_LOCK_SET(pr, PAGEFAULT);
+	vmassert(*old_eipptr == old_eip);
+	vmassert(old_eipptr != &old_eip);
 
-	if(pr->p_endpoint <= INIT_PROC_NR && !(pr->p_misc_flags & MF_FULLVM)) {
+#if 0
+	printf("kernel: pagefault in pr %d, addr 0x%lx, his cr3 0x%lx, actual cr3 0x%lx\n",
+		pr->p_endpoint, pagefaultcr2, pr->p_seg.p_cr3, read_cr3());
+#endif
+
+	if(pr->p_seg.p_cr3) {
+#if 0
+		vm_print(pr->p_seg.p_cr3);
+#endif
+		vmassert(pr->p_seg.p_cr3 == read_cr3());
+	} else {
+		u32_t cr3;
+		int u = 0;
+		if(!intr_disabled()) {
+			lock;
+			u = 1;
+		}
+		cr3 = read_cr3();
+		vmassert(ptproc);
+		if(ptproc->p_seg.p_cr3 != cr3) {
+			util_stacktrace();
+			printf("cr3 wrong in pagefault; value 0x%lx, ptproc %s / %d, his cr3 0x%lx, pr %s / %d\n",
+				cr3,
+				ptproc->p_name, ptproc->p_endpoint,
+				ptproc->p_seg.p_cr3, 
+				pr->p_name, pr->p_endpoint);
+			ser_dump_proc();
+			vm_print(cr3);
+			vm_print(ptproc->p_seg.p_cr3);
+		}
+		if(u) {
+			unlock;
+		}
+	}
+
+	test_eip = k_reenter ? old_eip : pr->p_reg.pc;
+
+	in_physcopy = (test_eip > (vir_bytes) phys_copy) &&
+	   (test_eip < (vir_bytes) phys_copy_fault);
+
+	if((k_reenter || iskernelp(pr)) &&
+		catch_pagefaults && in_physcopy) {
+#if 0
+		printf("pf caught! addr 0x%lx\n", pagefaultcr2);
+#endif
+		*old_eipptr = phys_copy_fault;
+		*old_eaxptr = pagefaultcr2;
+
+		return;
+	}
+
+	/* System processes that don't have their own page table can't
+	 * have page faults. VM does have its own page table but also
+	 * can't have page faults (because VM has to handle them).
+	 */
+	if(k_reenter || (pr->p_endpoint <= INIT_PROC_NR &&
+	 !(pr->p_misc_flags & MF_FULLVM)) || pr->p_endpoint == VM_PROC_NR) {
 		/* Page fault we can't / don't want to
 		 * handle.
 		 */
-		kprintf("pagefault for process %d ('%s'), pc = 0x%x, addr = 0x%x, flags = 0x%x\n",
+		kprintf("pagefault for process %d ('%s'), pc = 0x%x, addr = 0x%x, flags = 0x%x, k_reenter %d\n",
 			pr->p_endpoint, pr->p_name, pr->p_reg.pc,
-			pagefault_cr2, trap_errno);
+			pagefaultcr2, trap_errno, k_reenter);
 		proc_stacktrace(pr);
   		minix_panic("page fault in system process", pr->p_endpoint);
 
 		return;
 	}
 
+	/* Don't schedule this process until pagefault is handled. */
+	vmassert(pr->p_seg.p_cr3 == read_cr3());
+	vmassert(!RTS_ISSET(pr, PAGEFAULT));
+	RTS_LOCK_SET(pr, PAGEFAULT);
+
 	/* Save pagefault details, suspend process,
 	 * add process to pagefault chain,
 	 * and tell VM there is a pagefault to be
 	 * handled.
 	 */
-	pr->p_pagefault.pf_virtual = pagefault_cr2;
+	pr->p_pagefault.pf_virtual = pagefaultcr2;
 	pr->p_pagefault.pf_flags = trap_errno;
 	pr->p_nextpagefault = pagefaults;
 	pagefaults = pr;
+		
 	lock_notify(HARDWARE, VM_PROC_NR);
 
-	pagefault_count = 0;
-
-#if 0
-	kprintf("pagefault for process %d ('%s'), pc = 0x%x\n",
-			pr->p_endpoint, pr->p_name, pr->p_reg.pc);
-	proc_stacktrace(pr);
-#endif
-
 	return;
 }
 
 /*===========================================================================*
  *				exception				     *
  *===========================================================================*/
-PUBLIC void exception(vec_nr, trap_errno, old_eip, old_cs, old_eflags)
+PUBLIC void exception(vec_nr, trap_errno, old_eip, old_cs, old_eflags,
+	old_eipptr, old_eaxptr, pagefaultcr2)
 unsigned vec_nr;
 u32_t trap_errno;
 u32_t old_eip;
 U16_t old_cs;
 u32_t old_eflags;
+u32_t *old_eipptr;
+u32_t *old_eaxptr;
+u32_t pagefaultcr2;
 {
 /* An exception or unexpected interrupt has occurred. */
 
@@ -108,13 +166,6 @@ struct proc *t;
   register struct ex_s *ep;
   struct proc *saved_proc;
 
-#if DEBUG_SCHED_CHECK
-  for (t = BEG_PROC_ADDR; t < END_PROC_ADDR; ++t) {
-	if(t->p_magic != PMAGIC)
-		kprintf("entry %d broken\n", t->p_nr);
-  }
-#endif
-
   /* Save proc_ptr, because it may be changed by debug statements. */
   saved_proc = proc_ptr;	
   
@@ -126,7 +177,8 @@ struct proc *t;
   }
 
   if(vec_nr == PAGE_FAULT_VECTOR) {
-		pagefault(saved_proc, trap_errno);
+	pagefault(old_eip, saved_proc, trap_errno,
+		old_eipptr, old_eaxptr, pagefaultcr2);
 	return;
   }
 
@@ -184,18 +236,21 @@ PUBLIC void proc_stacktrace(struct proc *proc)
 
 	v_bp = proc->p_reg.fp;
 
-	kprintf("%8.8s %6d 0x%lx ",
+	kprintf("%-8.8s %6d 0x%lx ",
 		proc->p_name, proc->p_endpoint, proc->p_reg.pc);
 
 	while(v_bp) {
-	        if(data_copy(proc->p_endpoint, v_bp,
-			SYSTEM, (vir_bytes) &v_hbp, sizeof(v_hbp)) != OK) {
+
+#define PRCOPY(pr, pv, v, n) \
+  (iskernelp(pr) ? (memcpy((char *) v, (char *) pv, n), OK) : \
+     data_copy(pr->p_endpoint, pv, SYSTEM, (vir_bytes) (v), n))
+
+	        if(PRCOPY(proc, v_bp, &v_hbp, sizeof(v_hbp)) != OK) {
 			kprintf("(v_bp 0x%lx ?)", v_bp);
 			break;
 		}
-		if(data_copy(proc->p_endpoint, v_bp + sizeof(v_pc),
-			SYSTEM, (vir_bytes) &v_pc, sizeof(v_pc)) != OK) {
-			kprintf("(v_pc 0x%lx ?)", v_pc);
+		if(PRCOPY(proc, v_bp + sizeof(v_pc), &v_pc, sizeof(v_pc)) != OK) {
+			kprintf("(v_pc 0x%lx ?)", v_bp + sizeof(v_pc));
 			break;
 		}
 		kprintf("0x%lx ", (unsigned long) v_pc);

kernel/arch/i386/include/archconst.h

@@ -136,5 +136,6 @@
 #define IOPL_MASK 0x003000
 
 #define vir2phys(vir)   (kinfo.data_base + (vir_bytes) (vir))
+#define phys2vir(ph)   ((vir_bytes) (ph) - kinfo.data_base)
 
 #endif /* _I386_ACONST_H */

kernel/arch/i386/include/archtypes.h

@@ -68,5 +68,7 @@ struct pagefault
 	u32_t   pf_flags;       /* Pagefault flags on stack. */
 };
 
+#define INMEMORY(p) (!p->p_seg.p_cr3 || ptproc == p)
+
 #endif /* #ifndef _I386_TYPES_H */
 

kernel/arch/i386/klib386.s

@@ -8,7 +8,6 @@
 #include <ibm/interrupt.h>
 #include <archconst.h>
 #include "../../const.h"
-#include "vm.h"
 #include "sconst.h"
 
 ! This file contains a number of assembly code utility routines needed by the
@@ -28,6 +27,7 @@
 .define	_intr_unmask	! enable an irq at the 8259 controller
 .define	_intr_mask	! disable an irq
 .define	_phys_copy	! copy data from anywhere to anywhere in memory
+.define	_phys_copy_fault! phys_copy pagefault
 .define	_phys_memset	! write pattern anywhere in memory
 .define	_mem_rdw	! copy one word from [segment:offset]
 .define	_reset		! reset the system
@@ -35,13 +35,12 @@
 .define	_level0		! call a function at level 0
 .define	_read_cpu_flags	! read the cpu flags
 .define	_read_cr0	! read cr0
-.define	_write_cr3	! write cr3
-.define _last_cr3
+.define	_getcr3val
 .define	_write_cr0	! write a value in cr0
 .define	_read_cr4
+.define	_thecr3
 .define	_write_cr4
-
-.define	_kernel_cr3	
+.define	_catch_pagefaults
 
 ! The routines only guarantee to preserve the registers the C compiler
 ! expects to be preserved (ebx, esi, edi, ebp, esp, segment registers, and
@@ -156,55 +155,6 @@ csinit:	mov	eax, DS_SELECTOR
 	ret
 
 
-!*===========================================================================*
-!*				cp_mess					     *
-!*===========================================================================*
-! PUBLIC void cp_mess(int src, phys_clicks src_clicks, vir_bytes src_offset,
-!		      phys_clicks dst_clicks, vir_bytes dst_offset);
-! This routine makes a fast copy of a message from anywhere in the address
-! space to anywhere else.  It also copies the source address provided as a
-! parameter to the call into the first word of the destination message.
-!
-! Note that the message size, "Msize" is in DWORDS (not bytes) and must be set
-! correctly.  Changing the definition of message in the type file and not
-! changing it here will lead to total disaster.
-!
-!CM_ARGS	=	4 + 4 + 4 + 4 + 4	! 4 + 4 + 4 + 4 + 4
-!!		es  ds edi esi eip	proc scl sof dcl dof
-!
-!	.align	16
-!_cp_mess:
-!	cld
-!	push	esi
-!	push	edi
-!	push	ds
-!	push	es
-!
-!	mov	eax, FLAT_DS_SELECTOR
-!	mov	ds, ax
-!	mov	es, ax
-!
-!	mov	esi, CM_ARGS+4(esp)		! src clicks
-!	shl	esi, CLICK_SHIFT
-!	add	esi, CM_ARGS+4+4(esp)		! src offset
-!	mov	edi, CM_ARGS+4+4+4(esp)		! dst clicks
-!	shl	edi, CLICK_SHIFT
-!	add	edi, CM_ARGS+4+4+4+4(esp)	! dst offset
-!
-!	mov	eax, CM_ARGS(esp)	! process number of sender
-!	stos				! copy number of sender to dest message
-!	add	esi, 4			! do not copy first word
-!	mov	ecx, Msize - 1		! remember, first word does not count
-!	rep
-!	movs				! copy the message
-!
-!	pop	es
-!	pop	ds
-!	pop	edi
-!	pop	esi
-!	ret				! that is all folks!
-!
-
 !*===========================================================================*
 !*				exit					     *
 !*===========================================================================*
@@ -236,8 +186,6 @@ _phys_insw:
 	push	edi
 	push	es
 
-	LOADKERNELCR3
-
 	mov	ecx, FLAT_DS_SELECTOR
 	mov	es, cx
 	mov	edx, 8(ebp)		! port to read from
@@ -264,8 +212,6 @@ _phys_insb:
 	push	edi
 	push	es
 
-	LOADKERNELCR3
-
 	mov	ecx, FLAT_DS_SELECTOR
 	mov	es, cx
 	mov	edx, 8(ebp)		! port to read from
@@ -293,8 +239,6 @@ _phys_outsw:
 	push	esi
 	push	ds
 
-	LOADKERNELCR3
-
 	mov	ecx, FLAT_DS_SELECTOR
 	mov	ds, cx
 	mov	edx, 8(ebp)		! port to write to
@@ -322,8 +266,6 @@ _phys_outsb:
 	push	esi
 	push	ds
 
-	LOADKERNELCR3
-
 	mov	ecx, FLAT_DS_SELECTOR
 	mov	ds, cx
 	mov	edx, 8(ebp)		! port to write to
@@ -416,7 +358,7 @@ dis_already:
 !*===========================================================================*
 !*				phys_copy				     *
 !*===========================================================================*
-! PUBLIC void phys_copy(phys_bytes source, phys_bytes destination,
+! PUBLIC phys_bytes phys_copy(phys_bytes source, phys_bytes destination,
 !			phys_bytes bytecount);
 ! Copy a block of physical memory.
 
@@ -430,8 +372,6 @@ _phys_copy:
 	push	edi
 	push	es
 
-	LOADKERNELCR3
-
 	mov	eax, FLAT_DS_SELECTOR
 	mov	es, ax
 
@@ -457,6 +397,8 @@ pc_small:
 	rep
    eseg	movsb
 
+	mov	eax, 0			! 0 means: no fault
+_phys_copy_fault:			! kernel can send us here
 	pop	es
 	pop	edi
 	pop	esi
@@ -477,8 +419,6 @@ _phys_memset:
 	push	ebx
 	push	ds
 
-	LOADKERNELCR3
-
 	mov	esi, 8(ebp)
 	mov	eax, 16(ebp)
 	mov	ebx, FLAT_DS_SELECTOR
@@ -633,14 +573,13 @@ _write_cr4:
 	pop	ebp
 	ret
 
+
 !*===========================================================================*
-!*				write_cr3				*
+!*				getcr3val				*
 !*===========================================================================*
-! PUBLIC void write_cr3(unsigned long value);
-_write_cr3:
-	push    ebp
-	mov     ebp, esp
-	LOADCR3WITHEAX(0x22, 8(ebp))
-	pop     ebp
+! PUBLIC unsigned long getcr3val(void);
+_getcr3val:
+	mov	eax, cr3
+	mov	(_thecr3), eax
 	ret
 

kernel/arch/i386/mpx386.s

@@ -60,7 +60,6 @@ begbss:
 #include <ibm/interrupt.h>
 #include <archconst.h>
 #include "../../const.h"
-#include "vm.h"
 #include "sconst.h"
 
 /* Selected 386 tss offsets. */
@@ -74,9 +73,8 @@ begbss:
 
 .define	_restart
 .define	save
-.define	_kernel_cr3
-.define	_pagefault_cr2
-.define _pagefault_count
+.define _reload_cr3
+.define	_write_cr3	! write cr3
 
 .define errexception
 .define exception1
@@ -101,6 +99,8 @@ begbss:
 .define	_params_size
 .define _params_offset
 .define _mon_ds
+.define _schedcheck
+.define _dirtypde
 
 .define	_hwint00	! handlers for hardware interrupts
 .define	_hwint01
@@ -218,12 +218,6 @@ csinit:
 	ltr	ax
 	push	0			! set flags to known good state
 	popf				! esp, clear nested task and int enable
-#if VM_KERN_NOPAGEZERO
-	jmp	laststep
-
-.align I386_PAGE_SIZE
-laststep:
-#endif
 	jmp	_main			! main()
 
 
@@ -239,7 +233,6 @@ laststep:
 #define hwint_master(irq)	\
 	call	save			/* save interrupted process state */;\
 	push	(_irq_handlers+4*irq)	/* irq_handlers[irq]		  */;\
-	LOADCR3WITHEAX(irq, (_kernel_cr3))	/* switch to kernel page table	  */;\
 	call	_intr_handle		/* intr_handle(irq_handlers[irq]) */;\
 	pop	ecx							    ;\
 	cmp	(_irq_actids+4*irq), 0	/* interrupt still active?	  */;\
@@ -291,7 +284,6 @@ _hwint07:		! Interrupt routine for irq 7 (printer)
 #define hwint_slave(irq)	\
 	call	save			/* save interrupted process state */;\
 	push	(_irq_handlers+4*irq)	/* irq_handlers[irq]		  */;\
-	LOADCR3WITHEAX(irq, (_kernel_cr3))	/* switch to kernel page table	  */;\
 	call	_intr_handle		/* intr_handle(irq_handlers[irq]) */;\
 	pop	ecx							    ;\
 	cmp	(_irq_actids+4*irq), 0	/* interrupt still active?	  */;\
@@ -398,11 +390,9 @@ _p_s_call:
 	push	eax		! source / destination
 	push	ecx		! call number (ipc primitive to use)
 
-!	LOADCR3WITHEAX(0x20, (_kernel_cr3))
-
 	call	_sys_call	! sys_call(call_nr, src_dst, m_ptr, bit_map)
 				! caller is now explicitly in proc_ptr
-	mov	AXREG(esi), eax	! sys_call MUST PRESERVE si
+	mov	AXREG(esi), eax
 
 ! Fall into code to restart proc/task running.
 
@@ -413,14 +403,21 @@ _restart:
 
 ! Restart the current process or the next process if it is set. 
 
-	cmp	(_next_ptr), 0		! see if another process is scheduled
-	jz	0f
-	mov 	eax, (_next_ptr)
-	mov	(_proc_ptr), eax	! schedule new process 
-	mov	(_next_ptr), 0
-0:	mov	esp, (_proc_ptr)	! will assume P_STACKBASE == 0
+	cli
+	call	_schedcheck		! ask C function who we're running
+	mov	esp, (_proc_ptr)	! will assume P_STACKBASE == 0
 	lldt	P_LDT_SEL(esp)		! enable process' segment descriptors 
-	LOADCR3WITHEAX(0x21, P_CR3(esp))	! switch to process page table
+	cmp	P_CR3(esp), 0		! process does not have its own PT
+	jz	0f	
+	mov 	eax, P_CR3(esp)
+	cmp	eax, (loadedcr3)
+	jz	0f
+	mov	cr3, eax
+	mov	(loadedcr3), eax
+	mov	eax, (_proc_ptr)
+	mov	(_ptproc), eax
+	mov	(_dirtypde), 0
+0:
 	lea	eax, P_STACKTOP(esp)	! arrange for next interrupt
 	mov	(_tss+TSS3_S_SP0), eax	! to save state in process table
 restart1:
@@ -496,8 +493,7 @@ _page_fault:
 	push	PAGE_FAULT_VECTOR
 	push	eax
 	mov	eax, cr2
-sseg	mov	(_pagefault_cr2), eax
-sseg	inc	(_pagefault_count)
+sseg	mov	(pagefaultcr2), eax
 	pop	eax
 	jmp	errexception
 
@@ -526,19 +522,26 @@ errexception:
  sseg	pop	(ex_number)
  sseg	pop	(trap_errno)
 exception1:				! Common for all exceptions.
+ sseg	mov	(old_eax_ptr), esp	! where will eax be saved?
+ sseg	sub	(old_eax_ptr), PCREG-AXREG	! here
+
 	push	eax			! eax is scratch register
 
 	mov	eax, 0+4(esp)		! old eip
  sseg	mov	(old_eip), eax
+	mov	eax, esp
+	add	eax, 4
+ sseg	mov	(old_eip_ptr), eax
 	movzx	eax, 4+4(esp)		! old cs
  sseg	mov	(old_cs), eax
 	mov	eax, 8+4(esp)		! old eflags
  sseg	mov	(old_eflags), eax
 
-	LOADCR3WITHEAX(0x24, (_kernel_cr3))
-
 	pop	eax
 	call	save
+	push	(pagefaultcr2)
+	push	(old_eax_ptr)
+	push	(old_eip_ptr)
 	push	(old_eflags)
 	push	(old_cs)
 	push	(old_eip)
@@ -546,7 +549,38 @@ exception1:				! Common for all exceptions.
 	push	(ex_number)
 	call	_exception		! (ex_number, trap_errno, old_eip,
 					!	old_cs, old_eflags)
-	add	esp, 5*4
+	add	esp, 8*4
+	ret
+
+
+!*===========================================================================*
+!*				write_cr3				*
+!*===========================================================================*
+! PUBLIC void write_cr3(unsigned long value);
+_write_cr3:
+	push    ebp
+	mov     ebp, esp
+	mov	eax, 8(ebp)
+	cmp	eax, (loadedcr3)
+	jz	0f
+	mov	cr3, eax
+	mov	(loadedcr3), eax
+	mov	(_dirtypde), 0
+0:
+	pop     ebp
+	ret
+
+!*===========================================================================*
+!*				reload_cr3				*
+!*===========================================================================*
+! PUBLIC void reload_cr3(void);
+_reload_cr3:
+	push    ebp
+	mov     ebp, esp
+	mov	(_dirtypde), 0
+	mov	eax, cr3
+	mov	cr3, eax
+	pop     ebp
 	ret
 
 !*===========================================================================*
@@ -556,24 +590,12 @@ _level0_call:
 	call	save
 	jmp	(_level0_func)
 
-!*===========================================================================*
-!*				load_kernel_cr3				     *
-!*===========================================================================*
-.align 16
-_load_kernel_cr3:
- 	mov	eax, (_kernel_cr3)
- 	mov	cr3, eax
-	ret
-
 !*===========================================================================*
 !*				data					     *
 !*===========================================================================*
 
 .sect .rom	! Before the string table please
 	.data2	0x526F		! this must be the first data entry (magic #)
-#if VM_KERN_NOPAGEZERO
-.align I386_PAGE_SIZE
-#endif
 
 .sect .bss
 k_stack:
@@ -581,7 +603,11 @@ k_stack:
 k_stktop:			! top of kernel stack
 	.comm	ex_number, 4
 	.comm	trap_errno, 4
+	.comm	old_eip_ptr, 4
+	.comm	old_eax_ptr, 4
 	.comm	old_eip, 4
 	.comm	old_cs, 4
 	.comm	old_eflags, 4
+	.comm	pagefaultcr2, 4
+	.comm	loadedcr3, 4
 

kernel/arch/i386/protect.c

@@ -189,6 +189,11 @@ PUBLIC void prot_init(void)
 	{ level0_call, LEVEL0_VECTOR, TASK_PRIVILEGE },
   };
 
+  /* Click-round kernel. */
+  if(kinfo.data_base % CLICK_SIZE)
+	minix_panic("kinfo.data_base not aligned", NO_NUM);
+  kinfo.data_size = ((kinfo.data_size+CLICK_SIZE-1)/CLICK_SIZE) * CLICK_SIZE;
+
   /* Build gdt and idt pointers in GDT where the BIOS expects them. */
   dtp= (struct desctableptr_s *) &gdt[GDT_INDEX];
   * (u16_t *) dtp->limit = (sizeof gdt) - 1;
@@ -325,3 +330,36 @@ PUBLIC void alloc_segments(register struct proc *rp)
       rp->p_reg.ds = (DS_LDT_INDEX*DESC_SIZE) | TI | privilege;
 }
 
+/*===========================================================================*
+ *				prot_set_kern_seg_limit			     *
+ *===========================================================================*/
+PUBLIC int prot_set_kern_seg_limit(vir_bytes limit)
+{
+	struct proc *rp;
+	vir_bytes prev;
+	int orig_click;
+	int incr_clicks;
+
+	if(limit <= kinfo.data_base) {
+		kprintf("prot_set_kern_seg_limit: limit bogus\n");
+		return EINVAL;
+	}
+
+	/* Do actual increase. */
+	orig_click = kinfo.data_size / CLICK_SIZE;
+	kinfo.data_size = limit - kinfo.data_base;
+	incr_clicks = kinfo.data_size / CLICK_SIZE - orig_click;
+
+	prot_init();
+
+	/* Increase kernel processes too. */
+	for (rp = BEG_PROC_ADDR; rp < END_PROC_ADDR; ++rp) {
+		if (RTS_ISSET(rp, SLOT_FREE) || !iskernelp(rp))
+			continue;
+		rp->p_memmap[S].mem_len += incr_clicks;
+		alloc_segments(rp);
+	}
+
+	return OK;
+}
+

kernel/arch/i386/proto.h

@@ -49,11 +49,17 @@ _PROTOTYPE( void vir_insb, (u16_t port, struct proc *proc, u32_t vir, size_t cou
 _PROTOTYPE( void vir_outsb, (u16_t port, struct proc *proc, u32_t vir, size_t count));
 _PROTOTYPE( void vir_insw, (u16_t port, struct proc *proc, u32_t vir, size_t count));
 _PROTOTYPE( void vir_outsw, (u16_t port, struct proc *proc, u32_t vir, size_t count));
+_PROTOTYPE( void i386_updatepde, (int pde, u32_t val));
+_PROTOTYPE( void i386_freepde, (int pde));
+_PROTOTYPE( void getcr3val, (void));
+_PROTOTYPE( void switchedcr3, (void));
+_PROTOTYPE( void vm_set_cr3, (struct proc *));
 
 
 /* exception.c */
 _PROTOTYPE( void exception, (unsigned vec_nr, u32_t trap_errno,
-	u32_t old_eip, U16_t old_cs, u32_t old_eflags)			);
+	u32_t old_eip, U16_t old_cs, u32_t old_eflags,
+	u32_t *old_eip_ptr, u32_t *old_eax_ptr, u32_t pagefaultcr2)	);
 
 /* klib386.s */
 _PROTOTYPE( void level0, (void (*func)(void))                           );
@@ -70,7 +76,12 @@ _PROTOTYPE( void phys_insb, (U16_t port, phys_bytes buf, size_t count)  );
 _PROTOTYPE( void phys_insw, (U16_t port, phys_bytes buf, size_t count)  );
 _PROTOTYPE( void phys_outsb, (U16_t port, phys_bytes buf, size_t count) );
 _PROTOTYPE( void phys_outsw, (U16_t port, phys_bytes buf, size_t count) );
-_PROTOTYPE( void i386_invlpg, (U32_t addr) );
+_PROTOTYPE( void i386_invlpg_level0, (void) );
+_PROTOTYPE( int _memcpy_k, (void *dst, void *src, size_t n) );
+_PROTOTYPE( int _memcpy_k_fault, (void) );
+_PROTOTYPE( u32_t read_cr3, (void) );
+_PROTOTYPE( void reload_cr3, (void) );
+_PROTOTYPE( void phys_memset, (phys_bytes ph, u32_t c, phys_bytes bytes)	);
 
 /* protect.c */
 _PROTOTYPE( void prot_init, (void)                     			);
@@ -79,6 +90,7 @@ _PROTOTYPE( void init_codeseg, (struct segdesc_s *segdp, phys_bytes base,
 _PROTOTYPE( void init_dataseg, (struct segdesc_s *segdp, phys_bytes base,
                 vir_bytes size, int privilege)                          );
 _PROTOTYPE( void enable_iop, (struct proc *pp)                          );
+_PROTOTYPE( int prot_set_kern_seg_limit, (vir_bytes limit)             );
 
 /* functions defined in architecture-independent kernel source. */
 #include "../../proto.h"

kernel/arch/i386/system.c

@@ -14,11 +14,11 @@
 
 #include "proto.h"
 #include "../../proc.h"
+#include "../../debug.h"
 
 #define CR0_EM	0x0004		/* set to enable trap on any FP instruction */
 
 FORWARD _PROTOTYPE( void ser_debug, (int c));
-FORWARD _PROTOTYPE( void ser_dump_stats, (void));
 
 PUBLIC void arch_shutdown(int how)
 {
@@ -137,82 +137,123 @@ PUBLIC void do_ser_debug()
 	ser_debug(c);
 }
 
+PRIVATE void ser_dump_queues(void)
+{
+	int q;
+	for(q = 0; q < NR_SCHED_QUEUES; q++) {
+		struct proc *p;
+		if(rdy_head[q])	
+			printf("%2d: ", q);
+		for(p = rdy_head[q]; p; p = p->p_nextready) {
+			printf("%s / %d  ", p->p_name, p->p_endpoint);
+		}
+		printf("\n");
+	}
+
+}
+
 PRIVATE void ser_debug(int c)
 {
+	int u = 0;
+
 	do_serial_debug++;
-	kprintf("ser_debug: %d\n", c);
+	/* Disable interrupts so that we get a consistent state. */
+	if(!intr_disabled()) { lock; u = 1; };
+
 	switch(c)
 	{
 	case '1':
 		ser_dump_proc();
 		break;
 	case '2':
-		ser_dump_stats();
+		ser_dump_queues();
 		break;
+#if DEBUG_TRACE
+#define TOGGLECASE(ch, flag)				\
+	case ch: {					\
+		if(verboseflags & flag)	{		\
+			verboseflags &= ~flag;		\
+			printf("%s disabled\n", #flag);	\
+		} else {				\
+			verboseflags |= flag;		\
+			printf("%s enabled\n", #flag);	\
+		}					\
+		break;					\
+		}
+	TOGGLECASE('8', VF_SCHEDULING)
+	TOGGLECASE('9', VF_PICKPROC)
+#endif
 	}
 	do_serial_debug--;
+	if(u) { unlock; }
 }
 
+PRIVATE void printslot(struct proc *pp, int level)
+{
+	struct proc *depproc = NULL;
+	int dep = NONE;
+#define COL { int i; for(i = 0; i < level; i++) printf("> "); }
+
+	if(level >= NR_PROCS) {
+		kprintf("loop??\n");
+		return;
+	}
+
+	COL
+
+	kprintf("%d: %s %d prio %d/%d time %d/%d cr3 0x%lx rts %s misc %s",
+		proc_nr(pp), pp->p_name, pp->p_endpoint, 
+		pp->p_priority, pp->p_max_priority, pp->p_user_time,
+		pp->p_sys_time, pp->p_seg.p_cr3,
+		rtsflagstr(pp->p_rts_flags), miscflagstr(pp->p_misc_flags));
+
+	if(pp->p_rts_flags & SENDING) {
+		dep = pp->p_sendto_e;
+		kprintf(" to: ");
+	} else if(pp->p_rts_flags & RECEIVING) {
+		dep = pp->p_getfrom_e;
+		kprintf(" from: ");
+	}
+
+	if(dep != NONE) {
+		if(dep == ANY) {
+			kprintf(" ANY\n");
+		} else {
+			int procno;
+			if(!isokendpt(dep, &procno)) {
+				kprintf(" ??? %d\n", dep);
+			} else {
+				depproc = proc_addr(procno);
+				if(depproc->p_rts_flags & SLOT_FREE) {
+					kprintf(" empty slot %d???\n", procno);
+					depproc = NULL;
+				} else {
+					kprintf(" %s\n", depproc->p_name);
+				}
+			}
+		}
+	} else {
+		kprintf("\n");
+	}
+
+	COL
+	proc_stacktrace(pp);
+
+	if(depproc)
+		printslot(depproc, level+1);
+}
+
+
 PUBLIC void ser_dump_proc()
 {
 	struct proc *pp;
-	int u = 0;
-
-	/* Disable interrupts so that we get a consistent state. */
-	if(!intr_disabled()) { lock; u = 1; };
 
 	for (pp= BEG_PROC_ADDR; pp < END_PROC_ADDR; pp++)
 	{
 		if (pp->p_rts_flags & SLOT_FREE)
 			continue;
-		kprintf(
-	"%d: 0x%02x %s e %d src %d dst %d prio %d/%d time %d/%d EIP 0x%x\n",
-			proc_nr(pp),
-			pp->p_rts_flags, pp->p_name,
-			pp->p_endpoint, pp->p_getfrom_e, pp->p_sendto_e,
-			pp->p_priority, pp->p_max_priority,
-			pp->p_user_time, pp->p_sys_time, 
-			pp->p_reg.pc);
-		proc_stacktrace(pp);
+		printslot(pp, 0);
 	}
-
-	if(u) { unlock; }
-}
-
-PRIVATE void ser_dump_stats()
-{
-	kprintf("ipc_stats:\n");
-	kprintf("deadproc: %d\n", ipc_stats.deadproc);
-	kprintf("bad_endpoint: %d\n", ipc_stats.bad_endpoint);
-	kprintf("dst_not_allowed: %d\n", ipc_stats.dst_not_allowed);
-	kprintf("bad_call: %d\n", ipc_stats.bad_call);
-	kprintf("call_not_allowed: %d\n", ipc_stats.call_not_allowed);
-	kprintf("bad_buffer: %d\n", ipc_stats.bad_buffer);
-	kprintf("deadlock: %d\n", ipc_stats.deadlock);
-	kprintf("not_ready: %d\n", ipc_stats.not_ready);
-	kprintf("src_died: %d\n", ipc_stats.src_died);
-	kprintf("dst_died: %d\n", ipc_stats.dst_died);
-	kprintf("no_priv: %d\n", ipc_stats.no_priv);
-	kprintf("bad_size: %d\n", ipc_stats.bad_size);
-	kprintf("bad_senda: %d\n", ipc_stats.bad_senda);
-	if (ex64hi(ipc_stats.total))
-	{
-		kprintf("total: %x:%08x\n", ex64hi(ipc_stats.total),
-			ex64lo(ipc_stats.total));
-	}
-	else
-		kprintf("total: %u\n", ex64lo(ipc_stats.total));
-
-	kprintf("sys_stats:\n");
-	kprintf("bad_req: %d\n", sys_stats.bad_req);
-	kprintf("not_allowed: %d\n", sys_stats.not_allowed);
-	if (ex64hi(sys_stats.total))
-	{
-		kprintf("total: %x:%08x\n", ex64hi(sys_stats.total),
-			ex64lo(sys_stats.total));
-	}
-	else
-		kprintf("total: %u\n", ex64lo(sys_stats.total));
 }
 
 #if SPROFILE

kernel/arch/i386/vm.h

@@ -1,27 +0,0 @@
-
-.define _load_kernel_cr3
-.define _last_cr3
-
-#define LOADKERNELCR3			;\
-	inc	(_cr3switch)		;\
-	mov	eax,	(_kernel_cr3)	;\
-	cmp	(_last_cr3), eax	;\
-	jz	9f			;\
-	push	_load_kernel_cr3	;\
-	call	_level0			;\
-	pop	eax			;\
-	mov	eax,	(_kernel_cr3)	;\
-	mov	(_last_cr3), eax	;\
-	inc	(_cr3reload)		;\
-9:
-
-#define LOADCR3WITHEAX(type, newcr3)	;\
-sseg	inc	(_cr3switch)		;\
-sseg	mov	eax,	newcr3		;\
-sseg	cmp	(_last_cr3), eax	;\
-	jz	8f			;\
-	mov	cr3, eax		;\
-sseg	inc	(_cr3reload)		;\
-sseg	mov	(_last_cr3), eax	;\
-8:
-

kernel/clock.c

@@ -99,6 +99,7 @@ message *m_ptr;				/* pointer to request message */
 /* Despite its name, this routine is not called on every clock tick. It
  * is called on those clock ticks when a lot of work needs to be done.
  */
+  vmassert(!vm_running || (read_cr3() == ptproc->p_seg.p_cr3));
   
   /* A process used up a full quantum. The interrupt handler stored this
    * process in 'prev_ptr'.  First make sure that the process is not on the 
@@ -120,12 +121,17 @@ message *m_ptr;				/* pointer to request message */
       }
   }
 
+  vmassert(!vm_running || (read_cr3() == ptproc->p_seg.p_cr3));
   /* Check if a clock timer expired and run its watchdog function. */
   if (next_timeout <= realtime) {
+  vmassert(!vm_running || (read_cr3() == ptproc->p_seg.p_cr3));
   	tmrs_exptimers(&clock_timers, realtime, NULL);  	
+  vmassert(!vm_running || (read_cr3() == ptproc->p_seg.p_cr3));
 	next_timeout = (clock_timers == NULL) ?
 		 TMR_NEVER : clock_timers->tmr_exp_time;	
+  vmassert(!vm_running || (read_cr3() == ptproc->p_seg.p_cr3));
   }
+  vmassert(!vm_running || (read_cr3() == ptproc->p_seg.p_cr3));
 
   return;
 }
@@ -188,6 +194,8 @@ irq_hook_t *hook;
 
   if(minix_panicing) return;
 
+  vmassert(!vm_running || (read_cr3() == ptproc->p_seg.p_cr3));
+
   /* Get number of ticks and update realtime. */
   ticks = lost_ticks + 1;
   lost_ticks = 0;
@@ -223,6 +231,7 @@ irq_hook_t *hook;
 
   if (do_serial_debug)
 	do_ser_debug();
+  vmassert(!vm_running || (read_cr3() == ptproc->p_seg.p_cr3));
 
   return(1);					/* reenable interrupts */
 }

kernel/debug.c

@@ -25,6 +25,8 @@ check_runqueues_f(char *file, int line)
 	minix_panic("check_runqueues called with interrupts enabled", NO_NUM);
   }
 
+  FIXME("check_runqueues being done");
+
 #define MYPANIC(msg) {		\
 	kprintf("check_runqueues:%s:%d: %s\n", file, line, msg); \
 	minix_panic("check_runqueues failed", NO_NUM);	\
@@ -94,7 +96,9 @@ check_runqueues_f(char *file, int line)
   for (xp = BEG_PROC_ADDR; xp < END_PROC_ADDR; ++xp) {
 	if(xp->p_magic != PMAGIC) 
 		MYPANIC("p_magic wrong in proc table");
-	if (! isemptyp(xp) && xp->p_ready && ! xp->p_found) {
+	if (isemptyp(xp))
+		continue;
+	if(xp->p_ready && ! xp->p_found) {
 		kprintf("sched error: ready proc %d not on queue\n", xp->p_nr);
 		MYPANIC("ready proc not on scheduling queue");
 		if (l++ > MAX_LOOP) { MYPANIC("loop in debug.c?"); }
@@ -103,3 +107,43 @@ check_runqueues_f(char *file, int line)
 }
 
 #endif /* DEBUG_SCHED_CHECK */
+
+PUBLIC char *
+rtsflagstr(int flags)
+{
+	static char str[100];
+	str[0] = '\0';
+
+#define FLAG(n) if(flags & n) { strcat(str, #n " "); }
+
+	FLAG(SLOT_FREE);
+	FLAG(NO_PRIORITY);
+	FLAG(SENDING);
+	FLAG(RECEIVING);
+	FLAG(SIGNALED);
+	FLAG(SIG_PENDING);
+	FLAG(P_STOP);
+	FLAG(NO_PRIV);
+	FLAG(NO_ENDPOINT);
+	FLAG(VMINHIBIT);
+	FLAG(PAGEFAULT);
+	FLAG(VMREQUEST);
+	FLAG(VMREQTARGET);
+
+	return str;
+}
+
+PUBLIC char *
+miscflagstr(int flags)
+{
+	static char str[100];
+	str[0] = '\0';
+
+	FLAG(MF_REPLY_PEND);
+	FLAG(MF_ASYNMSG);
+	FLAG(MF_FULLVM);
+	FLAG(MF_DELIVERMSG);
+
+	return str;
+}
+

kernel/debug.h

@@ -8,6 +8,7 @@
  */
 
 #include <ansi.h>
+#include <minix/debug.h>
 #include "config.h"
 
 /* Enable prints such as
@@ -24,7 +25,46 @@
 #define DEBUG_TIME_LOCKS		1
 
 /* Runtime sanity checking. */
-#define DEBUG_VMASSERT			1
+#define DEBUG_VMASSERT			0
 #define DEBUG_SCHED_CHECK		0
+#define DEBUG_STACK_CHECK		0
+#define DEBUG_TRACE			0
+
+#if DEBUG_TRACE
+
+#define VF_SCHEDULING		(1L << 1)
+#define VF_PICKPROC		(1L << 2)
+
+#define TRACE(code, statement) if(verboseflags & code) { printf("%s:%d: ", __FILE__, __LINE__); statement }
+
+#else
+#define TRACE(code, statement)
+#endif
+
+#define NOREC_ENTER(varname) \
+	static int varname = 0;	\
+	int mustunlock = 0; \
+	if(!intr_disabled()) { lock; mustunlock = 1; } \
+	if(varname) {	\
+		minix_panic(#varname " recursive enter", __LINE__); \
+	} \
+	varname = 1;
+
+#define NOREC_RETURN(varname, v) do {	\
+	if(!varname)		\
+		minix_panic(#varname " flag off", __LINE__); \
+	if(!intr_disabled())	\
+		minix_panic(#varname " interrupts on", __LINE__); \
+	varname = 0;	\
+	if(mustunlock)	{ unlock;	} \
+	return v;	\
+	} while(0)
+
+#if DEBUG_VMASSERT
+#define vmassert(t) { \
+	if(!(t)) { minix_panic("vm: assert " #t " failed\n", __LINE__); } }
+#else
+#define vmassert(t) { }
+#endif
 
 #endif /* DEBUG_H */

kernel/glo.h

@@ -16,6 +16,7 @@
 #include <minix/config.h>
 #include <archtypes.h>
 #include "config.h"
+#include "debug.h"
 
 /* Variables relating to shutting down MINIX. */
 EXTERN char kernel_exception;		/* TRUE after system exceptions */
@@ -29,14 +30,13 @@ EXTERN struct k_randomness krandom;	/* gather kernel random information */
 EXTERN struct loadinfo kloadinfo;	/* status of load average */
 
 /* Process scheduling information and the kernel reentry count. */
-EXTERN struct proc *prev_ptr;	/* previously running process */
 EXTERN struct proc *proc_ptr;	/* pointer to currently running process */
 EXTERN struct proc *next_ptr;	/* next process to run after restart() */
+EXTERN struct proc *prev_ptr;	
 EXTERN struct proc *bill_ptr;	/* process to bill for clock ticks */
 EXTERN struct proc *vmrestart;  /* first process on vmrestart queue */
 EXTERN struct proc *vmrequest;  /* first process on vmrequest queue */
 EXTERN struct proc *pagefaults; /* first process on pagefault queue */
-EXTERN struct proc *softnotify;	/* first process on softnotify queue */
 EXTERN char k_reenter;		/* kernel reentry count (entry count less 1) */
 EXTERN unsigned lost_ticks;	/* clock ticks counted outside clock task */
 
@@ -47,32 +47,6 @@ EXTERN int irq_actids[NR_IRQ_VECTORS];		/* IRQ ID bits active */
 EXTERN int irq_use;				/* map of all in-use irq's */
 EXTERN u32_t system_hz;				/* HZ value */
 
-EXTERN struct ipc_stats
-{
-	unsigned long deadproc;
-	unsigned long bad_endpoint;
-	unsigned long dst_not_allowed;
-	unsigned long bad_call;
-	unsigned long call_not_allowed;
-	unsigned long bad_buffer;
-	unsigned long deadlock;
-	unsigned long not_ready;
-	unsigned long src_died;
-	unsigned long dst_died;
-	unsigned long no_priv;
-	unsigned long bad_size;
-	unsigned long bad_senda;
-	u64_t total;
-} ipc_stats;
-extern endpoint_t ipc_stats_target;
-
-EXTERN struct system_stats
-{
-	unsigned long bad_req;
-	unsigned long not_allowed;
-	u64_t total;
-} sys_stats;
-
 /* Miscellaneous. */
 EXTERN reg_t mon_ss, mon_sp;		/* boot monitor stack */
 EXTERN int mon_return;			/* true if we can return to monitor */
@@ -85,18 +59,14 @@ EXTERN char params_buffer[512];		/* boot monitor parameters */
 EXTERN int minix_panicing;
 EXTERN int locklevel;
 
-EXTERN unsigned long cr3switch;
-EXTERN unsigned long cr3reload;
+#if DEBUG_TRACE
+EXTERN int verboseflags;
+#endif
 
 /* VM */
-EXTERN phys_bytes vm_base;
-EXTERN phys_bytes vm_size;
-EXTERN phys_bytes vm_mem_high;
 EXTERN int vm_running;
-EXTERN int must_notify_vm;
-
-/* Verbose flags (debugging). */
-EXTERN int verbose_vm;
+EXTERN int catch_pagefaults;
+EXTERN struct proc *ptproc;
 
 /* Timing */
 EXTERN util_timingdata_t timingdata[TIMING_CATEGORIES];

kernel/main.c

@@ -17,6 +17,7 @@
 #include <minix/com.h>
 #include <minix/endpoint.h>
 #include "proc.h"
+#include "debug.h"
 
 /* Prototype declarations for PRIVATE functions. */
 FORWARD _PROTOTYPE( void announce, (void));	
@@ -158,6 +159,9 @@ PUBLIC void main()
 		rp->p_reg.sp -= sizeof(reg_t);
 	}
 
+	/* scheduling functions depend on proc_ptr pointing somewhere. */
+	if(!proc_ptr) proc_ptr = rp;
+
 	/* If this process has its own page table, VM will set the
 	 * PT up and manage it. VM will signal the kernel when it has
 	 * done this; until then, don't let it run.
@@ -185,6 +189,19 @@ PUBLIC void main()
    */
   bill_ptr = proc_addr(IDLE);	/* it has to point somewhere */
   announce();				/* print MINIX startup banner */
+/* Warnings for sanity checks that take time. These warnings are printed
+ * so it's a clear warning no full release should be done with them
+ * enabled.
+ */
+#if DEBUG_SCHED_CHECK
+  FIXME("DEBUG_SCHED_CHECK enabled");
+#endif
+#if DEBUG_VMASSERT
+  FIXME("DEBUG_VMASSERT enabled");
+#endif
+#if DEBUG_PROC_CHECK
+  FIXME("PROC check enabled");
+#endif
   restart();
 }
 

kernel/proc.c

@@ -8,8 +8,6 @@
  *
  *   lock_notify:     notify a process of a system event
  *   lock_send:	      send a message to a process
- *   lock_enqueue:    put a process on one of the scheduling queues 
- *   lock_dequeue:    remove a process from the scheduling queues
  *
  * Changes:
  *   Aug 19, 2005     rewrote scheduling code  (Jorrit N. Herder)
@@ -67,8 +65,10 @@ FORWARD _PROTOTYPE( int try_one, (struct proc *src_ptr, struct proc *dst_ptr));
 FORWARD _PROTOTYPE( void sched, (struct proc *rp, int *queue, int *front));
 FORWARD _PROTOTYPE( void pick_proc, (void));
 
-#define BuildMess(m_ptr, src, dst_ptr) \
-	(m_ptr)->m_source = proc_addr(src)->p_endpoint;		\
+#define PICK_ANY	1
+#define PICK_HIGHERONLY	2
+
+#define BuildNotifyMessage(m_ptr, src, dst_ptr) \
 	(m_ptr)->m_type = NOTIFY_FROM(src);				\
 	(m_ptr)->NOTIFY_TIMESTAMP = get_uptime();			\
 	switch (src) {							\
@@ -82,49 +82,86 @@ FORWARD _PROTOTYPE( void pick_proc, (void));
 		break;							\
 	}
 
-#define CopyMess(s,sp,sm,dp,dm) do { 			\
-	vir_bytes dstlin;				\
-	endpoint_t e = proc_addr(s)->p_endpoint;	\
-	struct vir_addr src, dst;			\
-	int r;						\
-	if((dstlin = umap_local((dp), D, (vir_bytes) dm, sizeof(message))) == 0){\
-		minix_panic("CopyMess: umap_local failed", __LINE__);	\
-	}						\
-			\
-	if(vm_running &&	\
-	 (r=vm_checkrange((dp), (dp), dstlin, sizeof(message), 1, 0)) != OK) { \
-		if(r != VMSUSPEND) 			\
-		  minix_panic("CopyMess: vm_checkrange error", __LINE__); \
-		(dp)->p_vmrequest.saved.msgcopy.dst = (dp);	\
-		(dp)->p_vmrequest.saved.msgcopy.dst_v = (vir_bytes) dm;	\
-  		if(data_copy((sp)->p_endpoint,	\
-			(vir_bytes) (sm), SYSTEM,	\
-			(vir_bytes) &(dp)->p_vmrequest.saved.msgcopy.msgbuf, \
-			sizeof(message)) != OK) {		\
-				minix_panic("CopyMess: data_copy failed", __LINE__);\
-			}				\
-			(dp)->p_vmrequest.saved.msgcopy.msgbuf.m_source = e; \
-			(dp)->p_vmrequest.type = VMSTYPE_MSGCOPY; \
-	} else 	{					\
-		src.proc_nr_e = (sp)->p_endpoint;		\
-		dst.proc_nr_e = (dp)->p_endpoint;		\
-		src.segment = dst.segment = D;			\
-		src.offset = (vir_bytes) (sm);			\
-		dst.offset = (vir_bytes) (dm);			\
-		if(virtual_copy(&src, &dst, sizeof(message)) != OK) {	\
-			kprintf("copymess: copy %d:%lx to %d:%lx failed\n",\
-				(sp)->p_endpoint, (sm), (dp)->p_endpoint, dm);\
-			minix_panic("CopyMess: virtual_copy (1) failed", __LINE__); \
-		}		\
-		src.proc_nr_e = SYSTEM;				\
-		src.offset = (vir_bytes) &e;			\
-		if(virtual_copy(&src, &dst, sizeof(e)) != OK) {		\
-			kprintf("copymess: copy %d:%lx to %d:%lx\n",	\
-				(sp)->p_endpoint, (sm), (dp)->p_endpoint, dm);\
-			minix_panic("CopyMess: virtual_copy (2) failed", __LINE__); \
-		}					\
-	}	\
-} while(0)
+/*===========================================================================*
+ *				QueueMess				     * 
+ *===========================================================================*/
+PRIVATE int QueueMess(endpoint_t ep, vir_bytes msg_lin, struct proc *dst)
+{
+	int k;
+	phys_bytes addr;
+	NOREC_ENTER(queuemess);
+	/* Queue a message from the src process (in memory) to the dst
+	 * process (using dst process table entry). Do actual copy to
+	 * kernel here; it's an error if the copy fails into kernel.
+	 */
+	vmassert(!(dst->p_misc_flags & MF_DELIVERMSG));	
+	vmassert(dst->p_delivermsg_lin);
+	vmassert(isokendpt(ep, &k));
+
+#if 0
+	if(INMEMORY(dst)) {
+		PHYS_COPY_CATCH(msg_lin, dst->p_delivermsg_lin,
+			sizeof(message), addr);
+		if(!addr) {
+			PHYS_COPY_CATCH(vir2phys(&ep), dst->p_delivermsg_lin,
+				sizeof(ep), addr);
+			if(!addr) {
+				NOREC_RETURN(queuemess, OK);
+			}
+		}
+	}
+#endif
+
+	PHYS_COPY_CATCH(msg_lin, vir2phys(&dst->p_delivermsg), sizeof(message), addr);
+	if(addr) {
+		NOREC_RETURN(queuemess, EFAULT);
+	}
+
+	dst->p_delivermsg.m_source = ep;
+	dst->p_misc_flags |= MF_DELIVERMSG;
+
+	NOREC_RETURN(queuemess, OK);
+}
+
+/*===========================================================================*
+ *				schedcheck				     * 
+ *===========================================================================*/
+PUBLIC void schedcheck(void)
+{
+	/* This function is called an instant before proc_ptr is
+	 * to be scheduled again.
+	 */
+  	NOREC_ENTER(schedch);
+	vmassert(intr_disabled());
+	if(next_ptr) {
+		proc_ptr = next_ptr;
+		next_ptr = NULL;
+	}
+	vmassert(proc_ptr);
+	vmassert(!proc_ptr->p_rts_flags);
+	while(proc_ptr->p_misc_flags & MF_DELIVERMSG) {
+		vmassert(!next_ptr);
+		vmassert(!proc_ptr->p_rts_flags);
+		TRACE(VF_SCHEDULING, printf("delivering to %s / %d\n",
+			proc_ptr->p_name, proc_ptr->p_endpoint););
+		if(delivermsg(proc_ptr) == VMSUSPEND) {
+			vmassert(next_ptr);
+			TRACE(VF_SCHEDULING, printf("suspending %s / %d\n",
+				proc_ptr->p_name, proc_ptr->p_endpoint););
+			vmassert(proc_ptr->p_rts_flags);
+			vmassert(next_ptr != proc_ptr);
+			proc_ptr = next_ptr;
+			vmassert(!proc_ptr->p_rts_flags);
+			next_ptr = NULL;
+		} 
+	}
+	TRACE(VF_SCHEDULING, printf("starting %s / %d\n",
+		proc_ptr->p_name, proc_ptr->p_endpoint););
+#if DEBUG_TRACE
+	proc_ptr->p_schedules++;
+#endif
+	NOREC_RETURN(schedch, );
+}
 
 /*===========================================================================*
  *				sys_call				     * 
@@ -145,9 +182,15 @@ long bit_map;			/* notification event set or flags */
   int result;					/* the system call's result */
   int src_dst_p;				/* Process slot number */
   size_t msg_size;
+  phys_bytes linaddr = 0;
 
-  if (caller_ptr->p_endpoint == ipc_stats_target)
-	ipc_stats.total= add64u(ipc_stats.total, 1);
+#if DEBUG_SCHED_CHECK
+  if(caller_ptr->p_misc_flags & MF_DELIVERMSG) {
+	kprintf("sys_call: MF_DELIVERMSG on for %s / %d\n",
+		caller_ptr->p_name, caller_ptr->p_endpoint);
+	minix_panic("MF_DELIVERMSG on", NO_NUM);
+  }
+#endif
 
 #if 0
   if(src_dst_e != 4 && src_dst_e != 5 &&
@@ -163,12 +206,10 @@ long bit_map;			/* notification event set or flags */
   }
 #endif
 
-#if 1
+#if DEBUG_SCHED_CHECK
   if (RTS_ISSET(caller_ptr, SLOT_FREE))
   {
 	kprintf("called by the dead?!?\n");
-	if (caller_ptr->p_endpoint == ipc_stats_target)
-		ipc_stats.deadproc++;
 	return EINVAL;
   }
 #endif
@@ -193,8 +234,6 @@ long bit_map;			/* notification event set or flags */
 		kprintf("sys_call: trap %d by %d with bad endpoint %d\n", 
 			call_nr, proc_nr(caller_ptr), src_dst_e);
 #endif
-  		if (caller_ptr->p_endpoint == ipc_stats_target)
-			ipc_stats.bad_endpoint++;
 		return EINVAL;
 	}
 	src_dst_p = src_dst_e;
@@ -214,8 +253,6 @@ long bit_map;			/* notification event set or flags */
 		kprintf("sys_call: trap %d by %d with bad endpoint %d\n", 
 			call_nr, proc_nr(caller_ptr), src_dst_e);
 #endif
-  		if (caller_ptr->p_endpoint == ipc_stats_target)
-			ipc_stats.bad_endpoint++;
 		return EDEADSRCDST;
 	}
 
@@ -233,8 +270,6 @@ long bit_map;			/* notification event set or flags */
 				call_nr, proc_nr(caller_ptr),
 				caller_ptr->p_name, src_dst_p);
 #endif
-			if (caller_ptr->p_endpoint == ipc_stats_target)
-				ipc_stats.dst_not_allowed++;
 			return(ECALLDENIED);	/* call denied by ipc mask */
 		}
 	}
@@ -245,10 +280,8 @@ long bit_map;			/* notification event set or flags */
 #if DEBUG_ENABLE_IPC_WARNINGS
 			kprintf(
 			"sys_call: ipc mask denied trap %d from %d to %d\n",
-				call_nr, proc_nr(caller_ptr), src_dst_p);
+				call_nr, caller_ptr->p_endpoint, src_dst_e);
 #endif
-			if (caller_ptr->p_endpoint == ipc_stats_target)
-				ipc_stats.dst_not_allowed++;
 			return(ECALLDENIED);	/* call denied by ipc mask */
 		}
 	}
@@ -261,8 +294,6 @@ long bit_map;			/* notification event set or flags */
       kprintf("sys_call: trap %d not allowed, caller %d, src_dst %d\n", 
           call_nr, proc_nr(caller_ptr), src_dst_p);
 #endif
-	if (caller_ptr->p_endpoint == ipc_stats_target)
-		ipc_stats.bad_call++;
 	return(ETRAPDENIED);		/* trap denied by mask or kernel */
   }
 
@@ -275,8 +306,6 @@ long bit_map;			/* notification event set or flags */
       kprintf("sys_call: trap %d not allowed, caller %d, src_dst %d\n", 
           call_nr, proc_nr(caller_ptr), src_dst_p);
 #endif
-	if (caller_ptr->p_endpoint == ipc_stats_target)
-		ipc_stats.call_not_allowed++;
 	return(ETRAPDENIED);		/* trap denied by mask or kernel */
   }
 
@@ -285,8 +314,6 @@ long bit_map;			/* notification event set or flags */
       kprintf("sys_call: trap %d not allowed, caller %d, src_dst %d\n", 
           call_nr, proc_nr(caller_ptr), src_dst_e);
 #endif
-	if (caller_ptr->p_endpoint == ipc_stats_target)
-		ipc_stats.call_not_allowed++;
 	return(ETRAPDENIED);		/* trap denied by mask or kernel */
   }
 
@@ -307,61 +334,6 @@ long bit_map;			/* notification event set or flags */
 	msg_size = sizeof(*m_ptr);
   }
 
-  /* If the call involves a message buffer, i.e., for SEND, SENDREC, 
-   * or RECEIVE, check the message pointer. This check allows a message to be 
-   * anywhere in data or stack or gap. It will have to be made more elaborate 
-   * for machines which don't have the gap mapped. 
-   *
-   * We use msg_size decided above.
-   */
-  if (call_nr == SEND || call_nr == SENDREC ||
-	call_nr == RECEIVE || call_nr == SENDA || call_nr == SENDNB) {
-	int r;
-	phys_bytes lin;
-
-	/* Map to linear address. */
-	if(msg_size > 0 && 
-		(lin = umap_local(caller_ptr, D, (vir_bytes) m_ptr, msg_size)) == 0) {
-		kprintf("umap_local failed for %s / %d on 0x%lx size %d\n",
-			caller_ptr->p_name, caller_ptr->p_endpoint,
-			m_ptr, msg_size);
-		return EFAULT;
-	}
-
-	/* Check if message pages in calling process are mapped.
-	 * We don't have to check the recipient if this is a send,
-	 * because this code will do that before its receive() starts.
-	 *
-	 * It is important the range is verified as _writable_, because
-	 * the kernel will want to write to the SENDA buffer in the future,
-	 * and those pages may not be shared between processes.
-	 */
-
-	if(vm_running && msg_size > 0 &&
-	 (r=vm_checkrange(caller_ptr, caller_ptr, lin, msg_size, 1, 0)) != OK) {
-		if(r != VMSUSPEND) {
-			kprintf("SYSTEM:sys_call:vm_checkrange: err %d\n", r);
-			return r;
-		}
-		
-		/* We can't go ahead with this call. Caller is suspended
-		 * and we have to save the state in its process struct.
-		 */
-		caller_ptr->p_vmrequest.saved.sys_call.call_nr = call_nr;
-		caller_ptr->p_vmrequest.saved.sys_call.m_ptr = m_ptr;
-		caller_ptr->p_vmrequest.saved.sys_call.src_dst_e = src_dst_e;
-		caller_ptr->p_vmrequest.saved.sys_call.bit_map = bit_map;
-		caller_ptr->p_vmrequest.type = VMSTYPE_SYS_CALL;
-
-		kprintf("SYSTEM: %s:%d: suspending call 0x%lx on ipc buffer 0x%lx length 0x%lx\n",
-			caller_ptr->p_name, caller_ptr->p_endpoint, call_nr, m_ptr, msg_size);
-
-		/* vm_checkrange() will have suspended caller with VMREQUEST. */
-		return OK;
-	}
-
-  } 
-
   /* Check for a possible deadlock for blocking SEND(REC) and RECEIVE. */
   if (call_nr == SEND || call_nr == SENDREC || call_nr == RECEIVE) {
       if (group_size = deadlock(call_nr, caller_ptr, src_dst_p)) {
@@ -369,8 +341,6 @@ long bit_map;			/* notification event set or flags */
           kprintf("sys_call: trap %d from %d to %d deadlocked, group size %d\n",
               call_nr, proc_nr(caller_ptr), src_dst_p, group_size);
 #endif
-	if (caller_ptr->p_endpoint == ipc_stats_target)
-		ipc_stats.deadlock++;
         return(ELOCKED);
       }
   }
@@ -386,7 +356,7 @@ long bit_map;			/* notification event set or flags */
   switch(call_nr) {
   case SENDREC:
 	/* A flag is set so that notifications cannot interrupt SENDREC. */
-	caller_ptr->p_misc_flags |= REPLY_PENDING;
+	caller_ptr->p_misc_flags |= MF_REPLY_PEND;
 	/* fall through */
   case SEND:			
 	result = mini_send(caller_ptr, src_dst_e, m_ptr, 0);
@@ -395,7 +365,7 @@ long bit_map;			/* notification event set or flags */
 	/* fall through for SENDREC */
   case RECEIVE:			
 	if (call_nr == RECEIVE)
-		caller_ptr->p_misc_flags &= ~REPLY_PENDING;
+		caller_ptr->p_misc_flags &= ~MF_REPLY_PEND;
 	result = mini_receive(caller_ptr, src_dst_e, m_ptr, 0);
 	break;
   case NOTIFY:
@@ -484,22 +454,6 @@ int src_dst;					/* src or dst process */
   return(0);					/* not a deadlock */
 }
 
-/*===========================================================================*
- *				sys_call_restart			     * 
- *===========================================================================*/
-PUBLIC void sys_call_restart(caller)
-struct proc *caller;
-{
-	int r;
-	kprintf("restarting sys_call code 0x%lx, "
-		"m_ptr 0x%lx, srcdst %d, bitmap 0x%lx, but not really\n",
-		caller->p_vmrequest.saved.sys_call.call_nr,
-		caller->p_vmrequest.saved.sys_call.m_ptr,
-		caller->p_vmrequest.saved.sys_call.src_dst_e,
-		caller->p_vmrequest.saved.sys_call.bit_map);
-	caller->p_reg.retreg = r;
-}
-
 /*===========================================================================*
  *				mini_send				     * 
  *===========================================================================*/
@@ -516,14 +470,19 @@ int flags;
   register struct proc *dst_ptr;
   register struct proc **xpp;
   int dst_p;
+  phys_bytes linaddr;
+  vir_bytes addr;
+  int r;
 
+  if(!(linaddr = umap_local(caller_ptr, D, (vir_bytes) m_ptr,
+	sizeof(message)))) {
+	return EFAULT;
+  }
   dst_p = _ENDPOINT_P(dst_e);
   dst_ptr = proc_addr(dst_p);
 
   if (RTS_ISSET(dst_ptr, NO_ENDPOINT))
   {
-	if (caller_ptr->p_endpoint == ipc_stats_target)
-		ipc_stats.dst_died++;
 	return EDSTDIED;
   }
 
@@ -532,18 +491,20 @@ int flags;
    */
   if (WILLRECEIVE(dst_ptr, caller_ptr->p_endpoint)) {
 	/* Destination is indeed waiting for this message. */
-	CopyMess(caller_ptr->p_nr, caller_ptr, m_ptr, dst_ptr,
-		 dst_ptr->p_messbuf);
+	vmassert(!(dst_ptr->p_misc_flags & MF_DELIVERMSG));	
+	if((r=QueueMess(caller_ptr->p_endpoint, linaddr, dst_ptr)) != OK)
+		return r;
 	RTS_UNSET(dst_ptr, RECEIVING);
   } else {
 	if(flags & NON_BLOCKING) {
-		if (caller_ptr->p_endpoint == ipc_stats_target)
-			ipc_stats.not_ready++;
 		return(ENOTREADY);
 	}
 
 	/* Destination is not waiting.  Block and dequeue caller. */
-	caller_ptr->p_messbuf = m_ptr;
+	PHYS_COPY_CATCH(linaddr, vir2phys(&caller_ptr->p_sendmsg),
+		sizeof(message), addr);
+
+	if(addr) { return EFAULT; }
 	RTS_SET(caller_ptr, SENDING);
 	caller_ptr->p_sendto_e = dst_e;
 
@@ -576,6 +537,18 @@ int flags;
   sys_map_t *map;
   bitchunk_t *chunk;
   int i, r, src_id, src_proc_nr, src_p;
+  phys_bytes linaddr;
+
+  vmassert(!(caller_ptr->p_misc_flags & MF_DELIVERMSG));
+
+  if(!(linaddr = umap_local(caller_ptr, D, (vir_bytes) m_ptr,
+	sizeof(message)))) {
+	return EFAULT;
+  }
+
+  /* This is where we want our message. */
+  caller_ptr->p_delivermsg_lin = linaddr;
+  caller_ptr->p_delivermsg_vir = (vir_bytes) m_ptr;
 
   if(src_e == ANY) src_p = ANY;
   else
@@ -583,8 +556,6 @@ int flags;
 	okendpt(src_e, &src_p);
 	if (RTS_ISSET(proc_addr(src_p), NO_ENDPOINT))
 	{
-		if (caller_ptr->p_endpoint == ipc_stats_target)
-			ipc_stats.src_died++;
 		return ESRCDIED;
 	}
   }
@@ -597,10 +568,11 @@ int flags;
   if (!RTS_ISSET(caller_ptr, SENDING)) {
 
     /* Check if there are pending notifications, except for SENDREC. */
-    if (! (caller_ptr->p_misc_flags & REPLY_PENDING)) {
+    if (! (caller_ptr->p_misc_flags & MF_REPLY_PEND)) {
 
         map = &priv(caller_ptr)->s_notify_pending;
         for (chunk=&map->chunk[0]; chunk<&map->chunk[NR_SYS_CHUNKS]; chunk++) {
+		endpoint_t hisep;
 
             /* Find a pending notification from the requested source. */ 
             if (! *chunk) continue; 			/* no bits in chunk */
@@ -617,8 +589,13 @@ int flags;
             *chunk &= ~(1 << i);			/* no longer pending */
 
             /* Found a suitable source, deliver the notification message. */
-	    BuildMess(&m, src_proc_nr, caller_ptr);	/* assemble message */
-            CopyMess(src_proc_nr, proc_addr(HARDWARE), &m, caller_ptr, m_ptr);
+	    BuildNotifyMessage(&m, src_proc_nr, caller_ptr);	/* assemble message */
+	    hisep = proc_addr(src_proc_nr)->p_endpoint;
+	    vmassert(!(caller_ptr->p_misc_flags & MF_DELIVERMSG));	
+	    vmassert(src_e == ANY || hisep == src_e);
+	    if((r=QueueMess(hisep, vir2phys(&m), caller_ptr)) != OK)  {
+		minix_panic("mini_receive: local QueueMess failed", NO_NUM);
+	    }
             return(OK);					/* report success */
         }
     }
@@ -627,20 +604,20 @@ int flags;
     xpp = &caller_ptr->p_caller_q;
     while (*xpp != NIL_PROC) {
         if (src_e == ANY || src_p == proc_nr(*xpp)) {
-#if 1
+#if DEBUG_SCHED_CHECK
 	    if (RTS_ISSET(*xpp, SLOT_FREE) || RTS_ISSET(*xpp, NO_ENDPOINT))
 	    {
 		kprintf("%d: receive from %d; found dead %d (%s)?\n",
 			caller_ptr->p_endpoint, src_e, (*xpp)->p_endpoint,
 			(*xpp)->p_name);
-		if (caller_ptr->p_endpoint == ipc_stats_target)
-			ipc_stats.deadproc++;
 		return EINVAL;
 	    }
 #endif
 
 	    /* Found acceptable message. Copy it and update status. */
-	    CopyMess((*xpp)->p_nr, *xpp, (*xpp)->p_messbuf, caller_ptr, m_ptr);
+  	    vmassert(!(caller_ptr->p_misc_flags & MF_DELIVERMSG));
+	    QueueMess((*xpp)->p_endpoint,
+		vir2phys(&(*xpp)->p_sendmsg), caller_ptr);
 	    RTS_UNSET(*xpp, SENDING);
             *xpp = (*xpp)->p_q_link;		/* remove from queue */
             return(OK);				/* report success */
@@ -659,7 +636,6 @@ int flags;
 	}
 	else
 	{
-		caller_ptr->p_messbuf = m_ptr;
 		r= try_async(caller_ptr);
 	}
 	if (r == OK)
@@ -672,12 +648,9 @@ int flags;
    */
   if ( ! (flags & NON_BLOCKING)) {
       caller_ptr->p_getfrom_e = src_e;		
-      caller_ptr->p_messbuf = m_ptr;
       RTS_SET(caller_ptr, RECEIVING);
       return(OK);
   } else {
-	if (caller_ptr->p_endpoint == ipc_stats_target)
-		ipc_stats.not_ready++;
 	return(ENOTREADY);
   }
 }
@@ -692,19 +665,22 @@ int dst;				/* which process to notify */
   register struct proc *dst_ptr = proc_addr(dst);
   int src_id;				/* source id for late delivery */
   message m;				/* the notification message */
+  int r;
 
   /* Check to see if target is blocked waiting for this message. A process 
    * can be both sending and receiving during a SENDREC system call.
    */
     if (WILLRECEIVE(dst_ptr, caller_ptr->p_endpoint) &&
-      ! (dst_ptr->p_misc_flags & REPLY_PENDING)) {
+      ! (dst_ptr->p_misc_flags & MF_REPLY_PEND)) {
       /* Destination is indeed waiting for a message. Assemble a notification 
        * message and deliver it. Copy from pseudo-source HARDWARE, since the
        * message is in the kernel's address space.
        */ 
-      BuildMess(&m, proc_nr(caller_ptr), dst_ptr);
-      CopyMess(proc_nr(caller_ptr), proc_addr(HARDWARE), &m, 
-          dst_ptr, dst_ptr->p_messbuf);
+      BuildNotifyMessage(&m, proc_nr(caller_ptr), dst_ptr);
+      vmassert(!(dst_ptr->p_misc_flags & MF_DELIVERMSG));
+      if((r=QueueMess(caller_ptr->p_endpoint, vir2phys(&m), dst_ptr)) != OK) {
+	minix_panic("mini_notify: local QueueMess failed", NO_NUM);
+      }
       RTS_UNSET(dst_ptr, RECEIVING);
       return(OK);
   } 
@@ -749,21 +725,20 @@ struct proc *caller_ptr;
 asynmsg_t *table;
 size_t size;
 {
-	int i, dst_p, done, do_notify;
+	int i, dst_p, done, do_notify, r;
 	unsigned flags;
 	struct proc *dst_ptr;
 	struct priv *privp;
 	message *m_ptr;
 	asynmsg_t tabent;
 	vir_bytes table_v = (vir_bytes) table;
+	vir_bytes linaddr;
 
 	privp= priv(caller_ptr);
 	if (!(privp->s_flags & SYS_PROC))
 	{
 		kprintf(
 		"mini_senda: warning caller has no privilege structure\n");
-		if (caller_ptr->p_endpoint == ipc_stats_target)
-			ipc_stats.no_priv++;
 		return EPERM;
 	}
 
@@ -777,6 +752,13 @@ size_t size;
 		return OK;
 	}
 
+	if(!(linaddr = umap_local(caller_ptr, D, (vir_bytes) table,
+		size * sizeof(*table)))) {
+		printf("mini_senda: umap_local failed; 0x%lx len 0x%lx\n",
+			table, size * sizeof(*table));
+		return EFAULT;
+	}
+
 	/* Limit size to something reasonable. An arbitrary choice is 16
 	 * times the number of process table entries.
 	 *
@@ -785,8 +767,6 @@ size_t size;
 	 */
 	if (size > 16*(NR_TASKS + NR_PROCS))
 	{
-		if (caller_ptr->p_endpoint == ipc_stats_target)
-			ipc_stats.bad_size++;
 		return EDOM;
 	}
 	
@@ -808,8 +788,6 @@ size_t size;
 		if (flags & ~(AMF_VALID|AMF_DONE|AMF_NOTIFY) ||
 			!(flags & AMF_VALID))
 		{
-			if (caller_ptr->p_endpoint == ipc_stats_target)
-				ipc_stats.bad_senda++;
 			return EINVAL;
 		}
 
@@ -833,11 +811,6 @@ size_t size;
 			continue;
 		}
 
-#if 0
-		kprintf("mini_senda: entry[%d]: flags 0x%x dst %d/%d\n",
-			i, tabent.flags, tabent.dst, dst_p);
-#endif
-
 		dst_ptr = proc_addr(dst_p);
 
 		/* NO_ENDPOINT should be removed */
@@ -866,12 +839,13 @@ size_t size;
 			m_ptr= &table[i].msg;	/* Note: pointer in the
 						 * caller's address space.
 						 */
-			CopyMess(caller_ptr->p_nr, caller_ptr, m_ptr, dst_ptr,
-				dst_ptr->p_messbuf);
-
+			/* Copy message from sender. */
+			tabent.result= QueueMess(caller_ptr->p_endpoint,
+				linaddr + (vir_bytes) &table[i].msg -
+					(vir_bytes) table, dst_ptr);
+			if(tabent.result == OK)
 				RTS_UNSET(dst_ptr, RECEIVING);
 
-			tabent.result= OK;
 			A_INSERT(i, result);
 			tabent.flags= flags | AMF_DONE;
 			A_INSERT(i, flags);
@@ -894,13 +868,6 @@ size_t size;
 	{
 		privp->s_asyntab= (vir_bytes)table;
 		privp->s_asynsize= size;
-#if 0
-		if(caller_ptr->p_endpoint > INIT_PROC_NR) {
-			kprintf("kernel: %s (%d) asynsend table at 0x%lx, %d\n", 
-				caller_ptr->p_name, caller_ptr->p_endpoint,
-				table, size);
-		}
-#endif
 	}
 	return OK;
 }
@@ -923,11 +890,8 @@ struct proc *caller_ptr;
 			continue;
 		if (privp->s_asynsize == 0)
 			continue;
-#if 0
-		kprintf("try_async: found asyntable for proc %d\n",
-			privp->s_proc_nr);
-#endif
 		src_ptr= proc_addr(privp->s_proc_nr);
+	  	vmassert(!(caller_ptr->p_misc_flags & MF_DELIVERMSG));
 		r= try_one(src_ptr, caller_ptr);
 		if (r == OK)
 			return r;
@@ -957,6 +921,7 @@ struct proc *dst_ptr;
 	asynmsg_t tabent;
 	vir_bytes table_v;
 	struct proc *caller_ptr;
+	int r;
 
 	privp= priv(src_ptr);
 	size= privp->s_asynsize;
@@ -986,8 +951,6 @@ struct proc *dst_ptr;
 		{
 			kprintf("try_one: bad bits in table\n");
 			privp->s_asynsize= 0;
-			if (src_ptr->p_endpoint == ipc_stats_target)
-				ipc_stats.bad_senda++;
 			return EINVAL;
 		}
 
@@ -1015,10 +978,11 @@ struct proc *dst_ptr;
 		m_ptr= &table_ptr[i].msg;	/* Note: pointer in the
 					 	 * caller's address space.
 					 	 */
-		CopyMess(src_ptr->p_nr, src_ptr, m_ptr, dst_ptr,
-			dst_ptr->p_messbuf);
+		A_RETRIEVE(i, msg);
+		r = QueueMess(src_ptr->p_endpoint, vir2phys(&tabent.msg),
+			dst_ptr);
 
-		tabent.result= OK;
+		tabent.result= r;
 		A_INSERT(i, result);
 		tabent.flags= flags | AMF_DONE;
 		A_INSERT(i, flags);
@@ -1066,43 +1030,6 @@ int dst_e;			/* (endpoint) who is to be notified */
   return(result);
 }
 
-/*===========================================================================*
- *				soft_notify				     *
- *===========================================================================*/
-PUBLIC int soft_notify(dst_e)
-int dst_e;			/* (endpoint) who is to be notified */
-{
-	int dst, u = 0;
-	struct proc *dstp, *sys = proc_addr(SYSTEM);
-
-/* Delayed interface to notify() from SYSTEM that is safe/easy to call
- * from more places than notify().
- */
-	if(!intr_disabled()) { lock; u = 1; }
-
-	{
-		if(!isokendpt(dst_e, &dst))
-			minix_panic("soft_notify to dead ep", dst_e);
-
-		dstp = proc_addr(dst);
-
-		if(!dstp->p_softnotified) {
-			dstp->next_soft_notify = softnotify;
-			softnotify = dstp;
-			dstp->p_softnotified = 1;
-	
-			if (RTS_ISSET(sys, RECEIVING)) {
-				sys->p_messbuf->m_source = SYSTEM;
-				RTS_UNSET(sys, RECEIVING);
-			}
-		}
-	}
-
-	if(u) { unlock; }
-
-	return OK;
-}
-
 /*===========================================================================*
  *				enqueue					     * 
  *===========================================================================*/
@@ -1117,15 +1044,19 @@ register struct proc *rp;	/* this process is now runnable */
   int q;	 				/* scheduling queue to use */
   int front;					/* add to front or back */
 
+  NOREC_ENTER(enqueuefunc);
+
 #if DEBUG_SCHED_CHECK
   if(!intr_disabled()) { minix_panic("enqueue with interrupts enabled", NO_NUM); }
-  CHECK_RUNQUEUES;
   if (rp->p_ready) minix_panic("enqueue already ready process", NO_NUM);
 #endif
 
   /* Determine where to insert to process. */
   sched(rp, &q, &front);
 
+  vmassert(q >= 0);
+  vmassert(q < IDLE_Q || rp->p_endpoint == IDLE);
+
   /* Now add the process to the queue. */
   if (rdy_head[q] == NIL_PROC) {		/* add to empty queue */
       rdy_head[q] = rdy_tail[q] = rp; 		/* create a new queue */
@@ -1141,19 +1072,29 @@ register struct proc *rp;	/* this process is now runnable */
       rp->p_nextready = NIL_PROC;		/* mark new end */
   }
 
-  /* Now select the next process to run, if there isn't a current
-   * process yet or current process isn't ready any more, or
-   * it's PREEMPTIBLE.
-   */
-  if(!proc_ptr || proc_ptr->p_rts_flags ||
-    (priv(proc_ptr)->s_flags & PREEMPTIBLE)) {
-     pick_proc();
-  }
-
 #if DEBUG_SCHED_CHECK
   rp->p_ready = 1;
   CHECK_RUNQUEUES;
 #endif
+
+  /* Now select the next process to run, if there isn't a current
+   * process yet or current process isn't ready any more, or
+   * it's PREEMPTIBLE.
+   */
+	vmassert(proc_ptr);
+#if 0
+  if(!proc_ptr || proc_ptr->p_rts_flags) 
+#else
+  if((proc_ptr->p_priority > rp->p_priority) &&
+   (priv(proc_ptr)->s_flags & PREEMPTIBLE)) 
+#endif
+     pick_proc();
+
+#if DEBUG_SCHED_CHECK
+  CHECK_RUNQUEUES;
+#endif
+
+  NOREC_RETURN(enqueuefunc, );
 }
 
 /*===========================================================================*
@@ -1170,14 +1111,17 @@ register struct proc *rp;	/* this process is no longer runnable */
   register struct proc **xpp;			/* iterate over queue */
   register struct proc *prev_xp;
 
+  NOREC_ENTER(dequeuefunc);
+
+#if DEBUG_STACK_CHECK
   /* Side-effect for kernel: check if the task's stack still is ok? */
   if (iskernelp(rp)) { 				
 	if (*priv(rp)->s_stack_guard != STACK_GUARD)
 		minix_panic("stack overrun by task", proc_nr(rp));
   }
+#endif
 
 #if DEBUG_SCHED_CHECK
-  CHECK_RUNQUEUES;
   if(!intr_disabled()) { minix_panic("dequeue with interrupts enabled", NO_NUM); }
   if (! rp->p_ready) minix_panic("dequeue() already unready process", NO_NUM);
 #endif
@@ -1193,6 +1137,11 @@ register struct proc *rp;	/* this process is no longer runnable */
           *xpp = (*xpp)->p_nextready;		/* replace with next chain */
           if (rp == rdy_tail[q])		/* queue tail removed */
               rdy_tail[q] = prev_xp;		/* set new tail */
+
+#if DEBUG_SCHED_CHECK
+  		rp->p_ready = 0;
+		  CHECK_RUNQUEUES;
+#endif
           if (rp == proc_ptr || rp == next_ptr)	/* active process removed */
               pick_proc();		/* pick new process to run */
           break;
@@ -1201,9 +1150,10 @@ register struct proc *rp;	/* this process is no longer runnable */
   }
 
 #if DEBUG_SCHED_CHECK
-  rp->p_ready = 0;
   CHECK_RUNQUEUES;
 #endif
+
+  NOREC_RETURN(dequeuefunc, );
 }
 
 /*===========================================================================*
@@ -1251,24 +1201,27 @@ PRIVATE void pick_proc()
   register struct proc *rp;			/* process to run */
   int q;				/* iterate over queues */
 
+  NOREC_ENTER(pick);
+
   /* Check each of the scheduling queues for ready processes. The number of
    * queues is defined in proc.h, and priorities are set in the task table.
    * The lowest queue contains IDLE, which is always ready.
    */
   for (q=0; q < NR_SCHED_QUEUES; q++) {	
-      if ( (rp = rdy_head[q]) != NIL_PROC) {
+	int found = 0;
+	if(!(rp = rdy_head[q])) {
+		TRACE(VF_PICKPROC, printf("queue %d empty\n", q););
+		continue;
+	}
+	TRACE(VF_PICKPROC, printf("found %s / %d on queue %d\n", 
+		rp->p_name, rp->p_endpoint, q););
 	next_ptr = rp;			/* run process 'rp' next */
-#if 0
-	  if(rp->p_endpoint != 4 && rp->p_endpoint != 5 && rp->p_endpoint != IDLE && rp->p_endpoint != SYSTEM)
-	  	kprintf("[run %s]",  rp->p_name);
-#endif
+	vmassert(!next_ptr->p_rts_flags);
 	if (priv(rp)->s_flags & BILLABLE)	 	
 		bill_ptr = rp;		/* bill for system time */
-          return;				 
+	NOREC_RETURN(pick, );
   }
 }
-  minix_panic("no ready process", NO_NUM);
-}
 
 /*===========================================================================*
  *				balance_queues				     *
@@ -1302,9 +1255,6 @@ timer_t *tp;					/* watchdog timer pointer */
 	  unlock;
       }
   }
-#if DEBUG
-  kprintf("ticks_added: %d\n", ticks_added);
-#endif
 
   /* Now schedule a new watchdog timer to balance the queues again.  The 
    * period depends on the total amount of quantum ticks added.
@@ -1328,37 +1278,6 @@ message *m_ptr;			/* pointer to message buffer */
   return(result);
 }
 
-/*===========================================================================*
- *				lock_enqueue				     *
- *===========================================================================*/
-PUBLIC void lock_enqueue(rp)
-struct proc *rp;		/* this process is now runnable */
-{
-/* Safe gateway to enqueue() for tasks. */
-  lock;
-  enqueue(rp);
-  unlock;
-}
-
-/*===========================================================================*
- *				lock_dequeue				     *
- *===========================================================================*/
-PUBLIC void lock_dequeue(rp)
-struct proc *rp;		/* this process is no longer runnable */
-{
-/* Safe gateway to dequeue() for tasks. */
-  if (k_reenter >= 0) {
-	/* We're in an exception or interrupt, so don't lock (and ... 
-	 * don't unlock).
-	 */
-	dequeue(rp);
-  } else {
-	lock;
-	dequeue(rp);
-	unlock;
-  }
-}
-
 /*===========================================================================*
  *				endpoint_lookup				     *
  *===========================================================================*/
@@ -1401,24 +1320,18 @@ int *p, fatalflag;
 	*p = _ENDPOINT_P(e);
 	if(!isokprocn(*p)) {
 #if DEBUG_ENABLE_IPC_WARNINGS
-#if 0
 		kprintf("kernel:%s:%d: bad endpoint %d: proc %d out of range\n",
 		file, line, e, *p);
-#endif
 #endif
 	} else if(isemptyn(*p)) {
 #if DEBUG_ENABLE_IPC_WARNINGS
-#if 0
 	kprintf("kernel:%s:%d: bad endpoint %d: proc %d empty\n", file, line, e, *p);
-#endif
 #endif
 	} else if(proc_addr(*p)->p_endpoint != e) {
 #if DEBUG_ENABLE_IPC_WARNINGS
-#if 0
 		kprintf("kernel:%s:%d: bad endpoint %d: proc %d has ept %d (generation %d vs. %d)\n", file, line,
 		e, *p, proc_addr(*p)->p_endpoint,
 		_ENDPOINT_G(e), _ENDPOINT_G(proc_addr(*p)->p_endpoint));
-#endif
 #endif
 	} else ok = 1;
 	if(!ok && fatalflag) {

kernel/proc.h

@@ -10,6 +10,7 @@
  * struct proc, be sure to change sconst.h to match.
  */
 #include <minix/com.h>
+#include <minix/portio.h>
 #include "const.h"
 #include "priv.h"
 
@@ -36,7 +37,6 @@ struct proc {
   struct proc *p_nextready;	/* pointer to next ready process */
   struct proc *p_caller_q;	/* head of list of procs wishing to send */
   struct proc *p_q_link;	/* link to next proc wishing to send */
-  message *p_messbuf;		/* pointer to passed message buffer */
   int p_getfrom_e;		/* from whom does process want to receive? */
   int p_sendto_e;		/* to whom does process want to send? */
 
@@ -46,6 +46,11 @@ struct proc {
 
   endpoint_t p_endpoint;	/* endpoint number, generation-aware */
 
+  message p_sendmsg;		/* Message from this process if SENDING */
+  message p_delivermsg;		/* Message for this process if MF_DELIVERMSG */
+  vir_bytes p_delivermsg_vir;	/* Virtual addr this proc wants message at */
+  vir_bytes p_delivermsg_lin;	/* Linear addr this proc wants message at */
+
   /* If handler functions detect a process wants to do something with
    * memory that isn't present, VM has to fix it. Until it has asked
    * what needs to be done and fixed it, save necessary state here.
@@ -57,28 +62,12 @@ struct proc {
 	struct proc	*nextrestart;	/* next in vmrestart chain */
 	struct proc	*nextrequestor;	/* next in vmrequest chain */
 #define VMSTYPE_SYS_NONE	0
-#define VMSTYPE_SYS_MESSAGE	1
-#define VMSTYPE_SYS_CALL	2
-#define VMSTYPE_MSGCOPY		3
+#define VMSTYPE_KERNELCALL	1
+#define VMSTYPE_DELIVERMSG	2
 	int		type;		/* suspended operation */
 	union {
 		/* VMSTYPE_SYS_MESSAGE */
 		message		reqmsg;	/* suspended request message */
-
-		/* VMSTYPE_SYS_CALL */
-		struct {
-                	int call_nr;
-                	message *m_ptr;
-                	int src_dst_e;
-                	long bit_map;
-		} sys_call;
-
-		/* VMSTYPE_MSGCOPY */
-		struct {
-			struct proc	*dst;
-			vir_bytes	dst_v;
-			message		msgbuf;
-		} msgcopy;
 	} saved;
 
 	/* Parameters of request to VM */
@@ -89,10 +78,9 @@ struct proc {
 	/* VM result when available */
 	int		vmresult;
 
-	/* Target gets this set. (But caller and target can be
-	 * the same, so we can't put this in the 'saved' union.)
-	*/
-	struct proc	*requestor;
+#if DEBUG_VMASSERT
+	char stacktrace[200];
+#endif
 
 	/* If the suspended operation is a sys_call, its details are
 	 * stored here.
@@ -107,6 +95,10 @@ struct proc {
 #define PMAGIC 0xC0FFEE1
   int p_magic;	/* check validity of proc pointers */
 #endif
+
+#if DEBUG_TRACE
+  int p_schedules;
+#endif
 };
 
 /* Bits for the runtime flags. A process is runnable iff p_rts_flags == 0. */
@@ -122,6 +114,7 @@ struct proc {
 #define VMINHIBIT       0x200	/* not scheduled until pagetable set by VM */
 #define PAGEFAULT       0x400	/* process has unhandled pagefault */
 #define VMREQUEST       0x800	/* originator of vm memory request */
+#define VMREQTARGET    0x1000	/* target of vm memory request */
 
 /* These runtime flags can be tested and manipulated by these macros. */
 
@@ -131,47 +124,60 @@ struct proc {
 /* Set flag and dequeue if the process was runnable. */
 #define RTS_SET(rp, f)							\
 	do {								\
+		vmassert(intr_disabled());				\
 		if(!(rp)->p_rts_flags) { dequeue(rp); }			\
 		(rp)->p_rts_flags |=  (f);				\
+		vmassert(intr_disabled());				\
 	} while(0)
 
 /* Clear flag and enqueue if the process was not runnable but is now. */
 #define RTS_UNSET(rp, f) 						\
 	do {								\
 		int rts;						\
+		vmassert(intr_disabled());				\
 		rts = (rp)->p_rts_flags;				\
 		(rp)->p_rts_flags &= ~(f);				\
 		if(rts && !(rp)->p_rts_flags) { enqueue(rp); }		\
+		vmassert(intr_disabled());				\
 	} while(0)
 
 /* Set flag and dequeue if the process was runnable. */
 #define RTS_LOCK_SET(rp, f)						\
 	do {								\
-		if(!(rp)->p_rts_flags) { lock_dequeue(rp); }		\
+		int u = 0;						\
+		if(!intr_disabled()) { u = 1; lock; }			\
+		if(!(rp)->p_rts_flags) { dequeue(rp); }			\
 		(rp)->p_rts_flags |=  (f);				\
+		if(u) { unlock;	}					\
 	} while(0)
 
 /* Clear flag and enqueue if the process was not runnable but is now. */
 #define RTS_LOCK_UNSET(rp, f) 						\
 	do {								\
 		int rts;						\
+		int u = 0;						\
+		if(!intr_disabled()) { u = 1; lock; }			\
 		rts = (rp)->p_rts_flags;				\
 		(rp)->p_rts_flags &= ~(f);				\
-		if(rts && !(rp)->p_rts_flags) { lock_enqueue(rp); }	\
+		if(rts && !(rp)->p_rts_flags) { enqueue(rp); }		\
+		if(u) { unlock;	}					\
 	} while(0)
 
 /* Set flags to this value. */
 #define RTS_LOCK_SETFLAGS(rp, f)					\
 	do {								\
-		if(!(rp)->p_rts_flags && (f)) { lock_dequeue(rp); }	\
+		int u = 0;						\
+		if(!intr_disabled()) { u = 1; lock; }			\
+		if(!(rp)->p_rts_flags && (f)) { dequeue(rp); }		\
 		(rp)->p_rts_flags = (f);				\
+		if(u) { unlock;	}					\
 	} while(0)
 
 /* Misc flags */
-#define REPLY_PENDING	0x01	/* reply to IPC_REQUEST is pending */
-#define MF_VM		0x08	/* process uses VM */
+#define MF_REPLY_PEND	0x01	/* reply to IPC_REQUEST is pending */
 #define MF_ASYNMSG	0x10	/* Asynchrous message pending */
 #define MF_FULLVM	0x20
+#define MF_DELIVERMSG	0x40	/* Copy message for him before running */
 
 /* Scheduling priorities for p_priority. Values must start at zero (highest
  * priority) and increment.  Priorities of the processes in the boot image 

kernel/proto.h

@@ -33,13 +33,11 @@ _PROTOTYPE( int sys_call, (int call_nr, int src_dst,
 					message *m_ptr, long bit_map)	);
 _PROTOTYPE( void sys_call_restart, (struct proc *caller)		);
 _PROTOTYPE( int lock_notify, (int src, int dst)				);
-_PROTOTYPE( int soft_notify, (int dst)					);
 _PROTOTYPE( int lock_send, (int dst, message *m_ptr)			);
-_PROTOTYPE( void lock_enqueue, (struct proc *rp)			);
-_PROTOTYPE( void lock_dequeue, (struct proc *rp)			);
 _PROTOTYPE( void enqueue, (struct proc *rp)				);
 _PROTOTYPE( void dequeue, (struct proc *rp)				);
 _PROTOTYPE( void balance_queues, (struct timer *tp)			);
+_PROTOTYPE( void schedcheck, (void)					);
 _PROTOTYPE( struct proc *endpoint_lookup, (endpoint_t ep)		);
 #if DEBUG_ENABLE_IPC_WARNINGS
 _PROTOTYPE( int isokendpt_f, (char *file, int line, endpoint_t e, int *p, int f));
@@ -87,6 +85,8 @@ _PROTOTYPE( void cons_seth, (int pos, int n)				);
 #define CHECK_RUNQUEUES check_runqueues_f(__FILE__, __LINE__)
 _PROTOTYPE( void check_runqueues_f, (char *file, int line) );
 #endif
+_PROTOTYPE( char *rtsflagstr, (int flags) );
+_PROTOTYPE( char *miscflagstr, (int flags) );
 
 /* system/do_safecopy.c */
 _PROTOTYPE( int verify_grant, (endpoint_t, endpoint_t, cp_grant_id_t, vir_bytes,
@@ -102,18 +102,21 @@ _PROTOTYPE( void stop_profile_clock, (void)				);
 #endif
 
 /* functions defined in architecture-dependent files. */
-_PROTOTYPE( void phys_copy, (phys_bytes source, phys_bytes dest,
+_PROTOTYPE( phys_bytes phys_copy, (phys_bytes source, phys_bytes dest,
                 phys_bytes count)                                       );
+_PROTOTYPE( void phys_copy_fault, (void));
 #define virtual_copy(src, dst, bytes) virtual_copy_f(src, dst, bytes, 0)
 #define virtual_copy_vmcheck(src, dst, bytes) virtual_copy_f(src, dst, bytes, 1)
 _PROTOTYPE( int virtual_copy_f, (struct vir_addr *src, struct vir_addr *dst, 
 				vir_bytes bytes, int vmcheck)		);
 _PROTOTYPE( int data_copy, (endpoint_t from, vir_bytes from_addr,
 	endpoint_t to, vir_bytes to_addr, size_t bytes));
+_PROTOTYPE( int data_copy_vmcheck, (endpoint_t from, vir_bytes from_addr,
+	endpoint_t to, vir_bytes to_addr, size_t bytes));
 #define data_copy_to(d, p, v, n) data_copy(SYSTEM, (d), (p), (v), (n));
 #define data_copy_from(d, p, v, n) data_copy((p), (v), SYSTEM, (d), (n));
 _PROTOTYPE( void alloc_segments, (struct proc *rp)                      );
-_PROTOTYPE( void vm_init, (void)                       			);
+_PROTOTYPE( void vm_init, (struct proc *first)        			);
 _PROTOTYPE( void vm_map_range, (u32_t base, u32_t size, u32_t offset)   );
 _PROTOTYPE( int vm_copy, (vir_bytes src, struct proc *srcproc,
 	vir_bytes dst, struct proc *dstproc, phys_bytes bytes));
@@ -126,7 +129,7 @@ _PROTOTYPE( phys_bytes umap_remote, (struct proc* rp, int seg,
 _PROTOTYPE( phys_bytes umap_virtual, (struct proc* rp, int seg,
         vir_bytes vir_addr, vir_bytes bytes)				);
 _PROTOTYPE( phys_bytes seg2phys, (U16_t)                                );
-_PROTOTYPE( void phys_memset, (phys_bytes source, unsigned long pattern,
+_PROTOTYPE( int vm_phys_memset, (phys_bytes source, u8_t pattern,
                 phys_bytes count)                                       );
 _PROTOTYPE( vir_bytes alloc_remote_segment, (u32_t *, segframe_t *,
         int, phys_bytes, vir_bytes, int));
@@ -160,5 +163,10 @@ _PROTOTYPE( int vm_checkrange, (struct proc *caller, struct proc *target,
 	vir_bytes start, vir_bytes length, int writeflag, int checkonly));
 _PROTOTYPE( void proc_stacktrace, (struct proc *proc)	         );
 _PROTOTYPE( int vm_lookup, (struct proc *proc, vir_bytes virtual, vir_bytes *result, u32_t *ptent));
+_PROTOTYPE( int vm_suspend, (struct proc *caller, struct proc *target,
+	phys_bytes lin, phys_bytes size, int wrflag, int type));
+_PROTOTYPE( int delivermsg, (struct proc *target));
+_PROTOTYPE( phys_bytes arch_switch_copymsg, (struct proc *rp, message *m,
+	phys_bytes lin));
 
 #endif /* PROTO_H */

kernel/system.c

@@ -38,6 +38,7 @@
 #include <sys/sigcontext.h>
 #include <minix/endpoint.h>
 #include <minix/safecopies.h>
+#include <minix/portio.h>
 #include <minix/u64.h>
 #include <sys/vm_i386.h>
 
@@ -56,7 +57,6 @@ char *callnames[NR_SYS_CALLS];
     call_vec[(call_nr-KERNEL_CALL)] = (handler)  
 
 FORWARD _PROTOTYPE( void initialize, (void));
-FORWARD _PROTOTYPE( void softnotify_check, (void));
 FORWARD _PROTOTYPE( struct proc *vmrestart_check, (message *));
 
 /*===========================================================================*
@@ -75,25 +75,17 @@ PUBLIC void sys_task()
   /* Initialize the system task. */
   initialize();
 
+
   while (TRUE) {
       struct proc *restarting;
 
       restarting = vmrestart_check(&m);
-      softnotify_check();
-	if(softnotify)
-		minix_panic("softnotify non-NULL before receive (1)", NO_NUM);
 
       if(!restarting) {
         int r;
 	/* Get work. Block and wait until a request message arrives. */
-	if(softnotify)
-		minix_panic("softnotify non-NULL before receive (2)", NO_NUM);
 	if((r=receive(ANY, &m)) != OK)
 		minix_panic("receive() failed", r);
-	if(m.m_source == SYSTEM)
-		continue;
-	if(softnotify)
-		minix_panic("softnotify non-NULL after receive", NO_NUM);
       } 
 
       sys_call_code = (unsigned) m.m_type;
@@ -102,37 +94,13 @@ PUBLIC void sys_task()
       okendpt(who_e, &who_p);
       caller_ptr = proc_addr(who_p);
 
-	if (caller_ptr->p_endpoint == ipc_stats_target)
-		sys_stats.total= add64u(sys_stats.total, 1);
-
       /* See if the caller made a valid request and try to handle it. */
       if (call_nr < 0 || call_nr >= NR_SYS_CALLS) {	/* check call number */
-#if DEBUG_ENABLE_IPC_WARNINGS
 	  kprintf("SYSTEM: illegal request %d from %d.\n",
 		call_nr,m.m_source);
-#endif
-	if (caller_ptr->p_endpoint == ipc_stats_target)
-		sys_stats.bad_req++;
 	  result = EBADREQUEST;			/* illegal message type */
       } 
       else if (!GET_BIT(priv(caller_ptr)->s_k_call_mask, call_nr)) {
-#if DEBUG_ENABLE_IPC_WARNINGS
-	static int curr= 0, limit= 100, extra= 20;
-
-	if (curr < limit+extra)
-	{
-		kprintf("SYSTEM: request %d from %d denied.\n",
-			call_nr, m.m_source);
-	} else if (curr == limit+extra)
-	{
-		kprintf("sys_task: no debug output for a while\n");
-	}
-	else if (curr == 2*limit-1)
-		limit *= 2;
-	curr++;
-#endif
-	if (caller_ptr->p_endpoint == ipc_stats_target)
-		sys_stats.not_allowed++;
 	  result = ECALLDENIED;			/* illegal message type */
       }
       else {
@@ -144,15 +112,18 @@ PUBLIC void sys_task()
 	 * until VM tells us it's allowed. VM has been notified
 	 * and we must wait for its reply to restart the call.
 	 */
+        vmassert(RTS_ISSET(caller_ptr, VMREQUEST));
+	vmassert(caller_ptr->p_vmrequest.type == VMSTYPE_KERNELCALL);
 	memcpy(&caller_ptr->p_vmrequest.saved.reqmsg, &m, sizeof(m));
-	caller_ptr->p_vmrequest.type = VMSTYPE_SYS_MESSAGE;
       } else if (result != EDONTREPLY) {
 	/* Send a reply, unless inhibited by a handler function.
 	 * Use the kernel function lock_send() to prevent a system
 	 * call trap.
 	 */
-		if(restarting)
-			RTS_LOCK_UNSET(restarting, VMREQUEST);
+		if(restarting) {
+        		vmassert(!RTS_ISSET(restarting, VMREQUEST));
+        		vmassert(!RTS_ISSET(restarting, VMREQTARGET));
+		}
 		m.m_type = result;		/* report status of call */
 		if(WILLRECEIVE(caller_ptr, SYSTEM)) {
 		  if (OK != (s=lock_send(m.m_source, &m))) {
@@ -220,7 +191,6 @@ PRIVATE void initialize(void)
   map(SYS_NEWMAP, do_newmap);		/* set up a process memory map */
   map(SYS_SEGCTL, do_segctl);		/* add segment and get selector */
   map(SYS_MEMSET, do_memset);		/* write char to memory area */
-  map(SYS_VM_SETBUF, do_vm_setbuf); 	/* PM passes buffer for page tables */
   map(SYS_VMCTL, do_vmctl);		/* various VM process settings */
 
   /* Copying. */
@@ -307,7 +277,7 @@ PUBLIC void send_sig(int proc_nr, int sig_nr)
 
   rp = proc_addr(proc_nr);
   sigaddset(&priv(rp)->s_sig_pending, sig_nr);
-  soft_notify(rp->p_endpoint); 
+  lock_notify(SYSTEM, rp->p_endpoint); 
 }
 
 /*===========================================================================*
@@ -424,7 +394,9 @@ register struct proc *rc;		/* slot of process to clean up */
 
   if(isemptyp(rc)) minix_panic("clear_proc: empty process", rc->p_endpoint);
 
-  if(rc->p_endpoint == PM_PROC_NR || rc->p_endpoint == VFS_PROC_NR) {
+#if 0
+  if(rc->p_endpoint == PM_PROC_NR || rc->p_endpoint == VFS_PROC_NR)
+  {
 	/* This test is great for debugging system processes dying,
 	 * but as this happens normally on reboot, not good permanent code.
 	 */
@@ -434,6 +406,7 @@ register struct proc *rc;		/* slot of process to clean up */
 	util_stacktrace();
 	minix_panic("clear_proc: system process died", rc->p_endpoint);
   }
+#endif
 
   /* Make sure that the exiting process is no longer scheduled. */
   RTS_LOCK_SET(rc, NO_ENDPOINT);
@@ -500,13 +473,6 @@ register struct proc *rc;		/* slot of process to clean up */
 #endif
       } 
   }
-
-  /* No pending soft notifies. */
-  for(np = softnotify; np; np = np->next_soft_notify) {
-    if(np == rc) {
-	minix_panic("dying proc was on next_soft_notify", np->p_endpoint);
-    }
-  }
 }
 
 /*===========================================================================*
@@ -540,28 +506,6 @@ int access;                     /* does grantee want to CPF_READ or _WRITE? */
         return umap_virtual(proc_addr(proc_nr), D, v_offset, bytes);
 } 
 
-/*===========================================================================*
- *                              softnotify_check                            *
- *===========================================================================*/
-PRIVATE void softnotify_check(void)
-{  
-	struct proc *np, *nextnp;
-
-	if(!softnotify) 
-		return;
-
-	for(np = softnotify; np; np = nextnp) {
-		if(!np->p_softnotified)
-			minix_panic("softnotify but no p_softnotified", NO_NUM);
-		lock_notify(SYSTEM, np->p_endpoint);
-		nextnp = np->next_soft_notify;
-		np->next_soft_notify = NULL;
-		np->p_softnotified = 0;
-	}
-
-	softnotify = NULL;
-}
-
 /*===========================================================================*
  *                              vmrestart_check                            *
  *===========================================================================*/
@@ -575,23 +519,18 @@ PRIVATE struct proc *vmrestart_check(message *m)
 	if(!(restarting = vmrestart))
 		return NULL;
 
-	if(restarting->p_rts_flags & SLOT_FREE)
-	   minix_panic("SYSTEM: VMREQUEST set for empty process", NO_NUM);
+	vmassert(!RTS_ISSET(restarting, SLOT_FREE));
+	vmassert(RTS_ISSET(restarting, VMREQUEST));
 
 	type = restarting->p_vmrequest.type;
 	restarting->p_vmrequest.type = VMSTYPE_SYS_NONE;
 	vmrestart = restarting->p_vmrequest.nextrestart;
 
-	if(!RTS_ISSET(restarting, VMREQUEST))
-	   minix_panic("SYSTEM: VMREQUEST not set for process on vmrestart queue",
-		restarting->p_endpoint);
-
 	switch(type) {
-		case VMSTYPE_SYS_MESSAGE:
+		case VMSTYPE_KERNELCALL:
 			memcpy(m, &restarting->p_vmrequest.saved.reqmsg, sizeof(*m));
-			if(m->m_source != restarting->p_endpoint)
-			   minix_panic("SYSTEM: vmrestart source doesn't match",
-				NO_NUM);
+			restarting->p_vmrequest.saved.reqmsg.m_source = NONE;
+			vmassert(m->m_source == restarting->p_endpoint);
 			/* Original caller could've disappeared in the meantime. */
 		        if(!isokendpt(m->m_source, &who_p)) {
 				kprintf("SYSTEM: ignoring call %d from dead %d\n",
@@ -610,26 +549,6 @@ PRIVATE struct proc *vmrestart_check(message *m)
 				}
 			}
 			return restarting;
-		case VMSTYPE_SYS_CALL:
-			kprintf("SYSTEM: restart sys_call\n");
-			/* Restarting a kernel trap. */
-			sys_call_restart(restarting);
-
-			/* Handled; restart system loop. */
-			return NULL;
-		case VMSTYPE_MSGCOPY:
-			/* Do delayed message copy. */
-			if((r=data_copy(SYSTEM,
-				(vir_bytes) &restarting->p_vmrequest.saved.msgcopy.msgbuf,
-				restarting->p_vmrequest.saved.msgcopy.dst->p_endpoint,
-				(vir_bytes) restarting->p_vmrequest.saved.msgcopy.dst_v,
-				sizeof(message))) != OK) {
-	   			minix_panic("SYSTEM: delayed msgcopy failed", r);
-			}
-			RTS_LOCK_UNSET(restarting, VMREQUEST);
-
-			/* Handled; restart system loop. */
-			return NULL;
 		default:
 	   		minix_panic("strange restart type", type);
 	}

kernel/system.h

@@ -91,9 +91,6 @@ _PROTOTYPE( int do_memset, (message *m_ptr) );
 #define do_memset do_unused
 #endif
 
-_PROTOTYPE( int do_vm_setbuf, (message *m_ptr) );
-_PROTOTYPE( int do_vm_map, (message *m_ptr) );
-
 _PROTOTYPE( int do_abort, (message *m_ptr) );
 #if ! USE_ABORT
 #define do_abort do_unused

kernel/system/Makefile

@@ -51,7 +51,6 @@ OBJECTS	= \
 	$(SYSTEM)(do_sigreturn.o) \
 	$(SYSTEM)(do_abort.o) \
 	$(SYSTEM)(do_getinfo.o) \
-	$(SYSTEM)(do_vm_setbuf.o) \
 	$(SYSTEM)(do_sprofile.o) \
 	$(SYSTEM)(do_cprofile.o) \
 	$(SYSTEM)(do_profbuf.o) \
@@ -162,9 +161,6 @@ $(SYSTEM)(do_vm.o):	do_vm.o
 do_vm.o:	do_vm.c
 	$(CC) do_vm.c
 
-$(SYSTEM)(do_vm_setbuf.o):	do_vm_setbuf.c
-	$(CC) do_vm_setbuf.c
-
 $(SYSTEM)(do_sprofile.o):       do_sprofile.c
 	$(CC) do_sprofile.c
 

kernel/system/do_devio.c

@@ -62,40 +62,18 @@ register message *m_ptr;	/* pointer to request message */
 			break;
 	}
 	if (i >= nr_io_range)
-	{
-		static int curr= 0, limit= 100, extra= 20;
-
-		if (curr < limit+extra)
 	{
 			kprintf("do_devio: port 0x%x (size %d) not allowed\n",
 				m_ptr->DIO_PORT, size);
-		} else if (curr == limit+extra)
-		{
-			kprintf("do_devio: no debug output for a while\n");
-		}
-		else if (curr == 2*limit-1)
-			limit *= 2;
-		curr++;
 		return EPERM;
 	}
     }
 
 doit:
     if (m_ptr->DIO_PORT & (size-1))
-    {
-	static int curr= 0, limit= 100, extra= 20;
-
-	if (curr < limit+extra)
     {
 		kprintf("do_devio: unaligned port 0x%x (size %d)\n",
 			m_ptr->DIO_PORT, size);
-	} else if (curr == limit+extra)
-	{
-		kprintf("do_devio: no debug output for a while\n");
-	}
-	else if (curr == 2*limit-1)
-		limit *= 2;
-	curr++;
 	return EPERM;
     }
 

kernel/system/do_exec.c

@@ -31,6 +31,11 @@ register message *m_ptr;	/* pointer to request message */
 
   rp = proc_addr(proc);
 
+  if(rp->p_misc_flags & MF_DELIVERMSG) {
+	rp->p_misc_flags &= ~MF_DELIVERMSG;
+	rp->p_delivermsg_lin = 0;
+  }
+
   /* Save command name for debugging, ps(1) output, etc. */
   if(data_copy(who_e, (vir_bytes) m_ptr->PR_NAME_PTR,
 	SYSTEM, (vir_bytes) rp->p_name, (phys_bytes) P_NAME_LEN - 1) != OK)

kernel/system/do_fork.c

@@ -7,6 +7,7 @@
  */
 
 #include "../system.h"
+#include "../vm.h"
 #include <signal.h>
 
 #include <minix/endpoint.h>
@@ -31,10 +32,25 @@ register message *m_ptr;	/* pointer to request message */
 
   if(!isokendpt(m_ptr->PR_ENDPT, &p_proc))
 	return EINVAL;
+
   rpp = proc_addr(p_proc);
   rpc = proc_addr(m_ptr->PR_SLOT);
   if (isemptyp(rpp) || ! isemptyp(rpc)) return(EINVAL);
 
+  vmassert(!(rpp->p_misc_flags & MF_DELIVERMSG));
+
+  /* needs to be receiving so we know where the message buffer is */
+  if(!RTS_ISSET(rpp, RECEIVING)) {
+	printf("kernel: fork not done synchronously?\n");
+	return EINVAL;
+  }
+
+  /* memory becomes readonly */
+  if (priv(rpp)->s_asynsize > 0) {
+	printf("kernel: process with waiting asynsend table can't fork\n");
+	return EINVAL;
+  }
+
   map_ptr= (struct mem_map *) m_ptr->PR_MEM_PTR;
 
   /* Copy parent 'proc' struct to child. And reinitialize some fields. */
@@ -75,9 +91,11 @@ register message *m_ptr;	/* pointer to request message */
 
   /* Calculate endpoint identifier, so caller knows what it is. */
   m_ptr->PR_ENDPT = rpc->p_endpoint;
+  m_ptr->PR_FORK_MSGADDR = (char *) rpp->p_delivermsg_vir;
 
   /* Install new map */
   r = newmap(rpc, map_ptr);
+  FIXLINMSG(rpc);
 
   /* Don't schedule process in VM mode until it has a new pagetable. */
   if(m_ptr->PR_FORK_FLAGS & PFF_VMINHIBIT) {

kernel/system/do_getinfo.c

@@ -28,9 +28,8 @@ register message *m_ptr;	/* pointer to request message */
  */
   size_t length;
   vir_bytes src_vir; 
-  int proc_nr, nr_e, nr;
+  int proc_nr, nr_e, nr, r;
   struct proc *caller;
-  phys_bytes ph;
   int wipe_rnd_bin = -1;
 
   caller = proc_addr(who_p);
@@ -67,19 +66,6 @@ register message *m_ptr;	/* pointer to request message */
         src_vir = (vir_bytes) irq_hooks;
         break;
     }
-    case GET_SCHEDINFO: {
-        /* This is slightly complicated because we need two data structures
-         * at once, otherwise the scheduling information may be incorrect.
-         * Copy the queue heads and fall through to copy the process table. 
-         */
-	if((ph=umap_local(caller, D, (vir_bytes) m_ptr->I_VAL_PTR2,length)) == 0)
-		return EFAULT;
-        length = sizeof(struct proc *) * NR_SCHED_QUEUES;
-  	CHECKRANGE_OR_SUSPEND(proc_addr(who_p), ph, length, 1);
-	data_copy(SYSTEM, (vir_bytes) rdy_head,
-		who_e, (vir_bytes) m_ptr->I_VAL_PTR2, length);
-        /* fall through to GET_PROCTAB */
-    }
     case GET_PROCTAB: {
         length = sizeof(struct proc) * (NR_PROCS + NR_TASKS);
         src_vir = (vir_bytes) proc;
@@ -174,15 +160,16 @@ register message *m_ptr;	/* pointer to request message */
 
   /* Try to make the actual copy for the requested data. */
   if (m_ptr->I_VAL_LEN > 0 && length > m_ptr->I_VAL_LEN) return (E2BIG);
-  if((ph=umap_local(caller, D, (vir_bytes) m_ptr->I_VAL_PTR,length)) == 0)
-	return EFAULT;
-  CHECKRANGE_OR_SUSPEND(caller, ph, length, 1);
-  if(data_copy(SYSTEM, src_vir, who_e, (vir_bytes) m_ptr->I_VAL_PTR, length) == OK) {
+  r = data_copy_vmcheck(SYSTEM, src_vir, who_e,
+	(vir_bytes) m_ptr->I_VAL_PTR, length);
+
+  if(r != OK) return r;
+
 	if(wipe_rnd_bin >= 0 && wipe_rnd_bin < RANDOM_SOURCES) {
 		krandom.bin[wipe_rnd_bin].r_size = 0;
 		krandom.bin[wipe_rnd_bin].r_next = 0;
 	}
-  }
+
   return(OK);
 }
 

kernel/system/do_memset.c

@@ -8,6 +8,7 @@
  */
 
 #include "../system.h"
+#include "../vm.h"
 
 #if USE_MEMSET
 
@@ -18,10 +19,8 @@ PUBLIC int do_memset(m_ptr)
 register message *m_ptr;
 {
 /* Handle sys_memset(). This writes a pattern into the specified memory. */
-  unsigned long p;
   unsigned char c = m_ptr->MEM_PATTERN;
-  p = c | (c << 8) | (c << 16) | (c << 24);
-  phys_memset((phys_bytes) m_ptr->MEM_PTR, p, (phys_bytes) m_ptr->MEM_COUNT);
+  vm_phys_memset((phys_bytes) m_ptr->MEM_PTR, c, (phys_bytes) m_ptr->MEM_COUNT);
   return(OK);
 }
 

kernel/system/do_safecopy.c

@@ -61,22 +61,11 @@ endpoint_t *e_granter;		/* new granter (magic grants) */
 	if(!HASGRANTTABLE(granter_proc)) return EPERM;
 
 	if(priv(granter_proc)->s_grant_entries <= grant) {
-		static int curr= 0, limit= 100, extra= 20;
-
-		if (curr < limit+extra)
-		{
 			kprintf(
 			"verify_grant: grant verify failed in ep %d proc %d: "
 			"grant %d out of range for table size %d\n",
 				granter, proc_nr, grant,
 				priv(granter_proc)->s_grant_entries);
-		} else if (curr == limit+extra)
-		{
-			kprintf("verify_grant: no debug output for a while\n");
-		}
-		else if (curr == 2*limit-1)
-			limit *= 2;
-		curr++;
 		return(EPERM);
 	}
 
@@ -219,23 +208,9 @@ int access;			/* CPF_READ for a copy from granter to grantee, CPF_WRITE
 	/* Verify permission exists. */
 	if((r=verify_grant(granter, grantee, grantid, bytes, access,
 	    g_offset, &v_offset, &new_granter)) != OK) {
-		static int curr= 0, limit= 100, extra= 20;
-
-		if (curr < limit+extra)
-		{
-#if 0
 			kprintf(
 		"grant %d verify to copy %d->%d by %d failed: err %d\n",
 				grantid, *src, *dst, grantee, r);
-#endif
-		} else if (curr == limit+extra)
-		{
-			kprintf(
-			"do_safecopy`safecopy: no debug output for a while\n");
-		}
-		else if (curr == 2*limit-1)
-			limit *= 2;
-		curr++;
 		return r;
 	}
 

kernel/system/do_sigsend.c

@@ -29,18 +29,13 @@ message *m_ptr;			/* pointer to request message */
   struct sigcontext sc, *scp;
   struct sigframe fr, *frp;
   int proc, r;
-  phys_bytes ph;
 
   if (!isokendpt(m_ptr->SIG_ENDPT, &proc)) return(EINVAL);
   if (iskerneln(proc)) return(EPERM);
   rp = proc_addr(proc);
 
-  ph = umap_local(proc_addr(who_p), D, (vir_bytes) m_ptr->SIG_CTXT_PTR, sizeof(struct sigmsg));
-  if(!ph) return EFAULT;
-  CHECKRANGE_OR_SUSPEND(proc_addr(who_p), ph, sizeof(struct sigmsg), 1);
-
   /* Get the sigmsg structure into our address space.  */
-  if((r=data_copy(who_e, (vir_bytes) m_ptr->SIG_CTXT_PTR,
+  if((r=data_copy_vmcheck(who_e, (vir_bytes) m_ptr->SIG_CTXT_PTR,
 	SYSTEM, (vir_bytes) &smsg, (phys_bytes) sizeof(struct sigmsg))) != OK)
 	return r;
 
@@ -59,12 +54,9 @@ message *m_ptr;			/* pointer to request message */
   sc.sc_flags = 0;	/* unused at this time */
   sc.sc_mask = smsg.sm_mask;
 
-  ph = umap_local(rp, D, (vir_bytes) scp, sizeof(struct sigcontext));
-  if(!ph) return EFAULT;
-  CHECKRANGE_OR_SUSPEND(rp, ph, sizeof(struct sigcontext), 1);
   /* Copy the sigcontext structure to the user's stack. */
-  if((r=data_copy(SYSTEM, (vir_bytes) &sc, m_ptr->SIG_ENDPT, (vir_bytes) scp,
-      (vir_bytes) sizeof(struct sigcontext))) != OK)
+  if((r=data_copy_vmcheck(SYSTEM, (vir_bytes) &sc, m_ptr->SIG_ENDPT,
+	(vir_bytes) scp, (vir_bytes) sizeof(struct sigcontext))) != OK)
       return r;
 
   /* Initialize the sigframe structure. */
@@ -78,11 +70,9 @@ message *m_ptr;			/* pointer to request message */
   fr.sf_signo = smsg.sm_signo;
   fr.sf_retadr = (void (*)()) smsg.sm_sigreturn;
 
-  ph = umap_local(rp, D, (vir_bytes) frp, sizeof(struct sigframe));
-  if(!ph) return EFAULT;
-  CHECKRANGE_OR_SUSPEND(rp, ph, sizeof(struct sigframe), 1);
   /* Copy the sigframe structure to the user's stack. */
-  if((r=data_copy(SYSTEM, (vir_bytes) &fr, m_ptr->SIG_ENDPT, (vir_bytes) frp, 
+  if((r=data_copy_vmcheck(SYSTEM, (vir_bytes) &fr,
+	m_ptr->SIG_ENDPT, (vir_bytes) frp, 
       (vir_bytes) sizeof(struct sigframe))) != OK)
       return r;
 

kernel/system/do_sysctl.c

@@ -16,7 +16,6 @@
 PUBLIC int do_sysctl(m_ptr)
 register message *m_ptr;	/* pointer to request message */
 {
-  phys_bytes ph;
   vir_bytes len, buf;
   static char mybuf[DIAG_BUFSIZE];
   struct proc *caller, *target;
@@ -33,10 +32,7 @@ register message *m_ptr;	/* pointer to request message */
 			caller->p_endpoint, len);
 		return EINVAL;
 	}
-	if((ph=umap_local(caller, D, buf, len)) == 0)
-		return EFAULT;
-	CHECKRANGE_OR_SUSPEND(caller, ph, len, 1);
-	if((s=data_copy(who_e, buf, SYSTEM, (vir_bytes) mybuf, len)) != OK) {
+	if((s=data_copy_vmcheck(who_e, buf, SYSTEM, (vir_bytes) mybuf, len)) != OK) {
 		kprintf("do_sysctl: diag for %d: len %d: copy failed: %d\n",
 			caller->p_endpoint, len, s);
 		return s;

kernel/system/do_umap.c

@@ -48,13 +48,11 @@ register message *m_ptr;	/* pointer to request message */
   case LOCAL_SEG:
       phys_addr = lin_addr = umap_local(targetpr, seg_index, offset, count); 
       if(!lin_addr) return EFAULT;
-      CHECKRANGE_OR_SUSPEND(targetpr, lin_addr, count, 1);
       naughty = 1;
       break;
   case REMOTE_SEG:
       phys_addr = lin_addr = umap_remote(targetpr, seg_index, offset, count); 
       if(!lin_addr) return EFAULT;
-      CHECKRANGE_OR_SUSPEND(targetpr, lin_addr, count, 1);
       naughty = 1;
       break;
   case GRANT_SEG:
@@ -93,7 +91,6 @@ register message *m_ptr;	/* pointer to request message */
 	kprintf("SYSTEM:do_umap: umap_local failed\n");
 	return EFAULT;
       }
-      CHECKRANGE_OR_SUSPEND(targetpr, lin_addr, count, 1);
       if(vm_lookup(targetpr, lin_addr, &phys_addr, NULL) != OK) {
 	kprintf("SYSTEM:do_umap: vm_lookup failed\n");
 	return EFAULT;

kernel/system/do_vm_setbuf.c

@@ -1,29 +0,0 @@
-/* The system call implemented in this file:
- *   m_type:	SYS_VM_SETBUF
- *
- * The parameters for this system call are:
- *    m4_l1:	Start of the buffer
- *    m4_l2:	Length of the buffer
- *    m4_l3:	End of main memory
- */
-#include "../system.h"
-
-#define VM_DEBUG 0		/* enable/ disable debug output */
-
-/*===========================================================================*
- *				do_vm_setbuf				     *
- *===========================================================================*/
-PUBLIC int do_vm_setbuf(m_ptr)
-message *m_ptr;			/* pointer to request message */
-{
-	vm_base= m_ptr->m4_l1;
-	vm_size= m_ptr->m4_l2;
-	vm_mem_high= m_ptr->m4_l3;
-
-#if VM_DEBUG
-	kprintf("do_vm_setbuf: got 0x%x @ 0x%x for 0x%x\n",
-		vm_size, vm_base, vm_mem_high);
-#endif
-
-	return OK;
-}

kernel/system/do_vmctl.c

@@ -13,6 +13,8 @@
 #include <minix/type.h>
 #include <minix/config.h>
 
+extern int verifyrange;
+
 /*===========================================================================*
  *				do_vmctl				     *
  *===========================================================================*/
@@ -21,12 +23,10 @@ register message *m_ptr;	/* pointer to request message */
 {
   int proc_nr, i;
   endpoint_t ep = m_ptr->SVMCTL_WHO;
-  struct proc *p, *rp;
+  struct proc *p, *rp, *target;
 
   if(ep == SELF) { ep = m_ptr->m_source; }
 
-  vm_init();
-
   if(!isokendpt(ep, &proc_nr)) {
 	kprintf("do_vmctl: unexpected endpoint %d from VM\n", ep);
 	return EINVAL;
@@ -42,14 +42,33 @@ register message *m_ptr;	/* pointer to request message */
 		/* Send VM the information about the memory request.  */
 		if(!(rp = vmrequest))
 			return ESRCH;
-		if(!RTS_ISSET(rp, VMREQUEST))
-			minix_panic("do_vmctl: no VMREQUEST set", NO_NUM);
+		vmassert(RTS_ISSET(rp, VMREQUEST));
+
+#if 0
+		printf("kernel: vm request sent by: %s / %d about %d; 0x%lx-0x%lx, wr %d, stack: %s ",
+			rp->p_name, rp->p_endpoint, rp->p_vmrequest.who,
+			rp->p_vmrequest.start,
+			rp->p_vmrequest.start + rp->p_vmrequest.length,
+			rp->p_vmrequest.writeflag, rp->p_vmrequest.stacktrace);
+		printf("type %d\n", rp->p_vmrequest.type);
+#endif
+
+#if DEBUG_VMASSERT
+  		okendpt(rp->p_vmrequest.who, &proc_nr);
+		target = proc_addr(proc_nr);
+		if(!RTS_ISSET(target, VMREQTARGET)) {
+			printf("set stack: %s\n", rp->p_vmrequest.stacktrace);
+			minix_panic("VMREQTARGET not set for target",
+				NO_NUM);
+		}
+#endif
 
 		/* Reply with request fields. */
 		m_ptr->SVMCTL_MRG_ADDR = (char *) rp->p_vmrequest.start;
 		m_ptr->SVMCTL_MRG_LEN = rp->p_vmrequest.length;
 		m_ptr->SVMCTL_MRG_WRITE = rp->p_vmrequest.writeflag;
 		m_ptr->SVMCTL_MRG_EP = rp->p_vmrequest.who;
+		m_ptr->SVMCTL_MRG_REQUESTOR = (void *) rp->p_endpoint;
 		rp->p_vmrequest.vmresult = VMSUSPEND;
 
 		/* Remove from request chain. */
@@ -57,46 +76,63 @@ register message *m_ptr;	/* pointer to request message */
 
 		return OK;
 	case VMCTL_MEMREQ_REPLY:
-		if(!(rp = p->p_vmrequest.requestor))
-			minix_panic("do_vmctl: no requestor set", ep);
-		p->p_vmrequest.requestor = NULL;
-		if(!RTS_ISSET(rp, VMREQUEST))
-			minix_panic("do_vmctl: no VMREQUEST set", ep);
-		if(rp->p_vmrequest.vmresult != VMSUSPEND)
-			minix_panic("do_vmctl: result not VMSUSPEND set",
-				rp->p_vmrequest.vmresult);
-		rp->p_vmrequest.vmresult = m_ptr->SVMCTL_VALUE;
-		if(rp->p_vmrequest.vmresult == VMSUSPEND)
-			minix_panic("VM returned VMSUSPEND?", NO_NUM);
-		if(rp->p_vmrequest.vmresult != OK)
+		vmassert(RTS_ISSET(p, VMREQUEST));
+		vmassert(p->p_vmrequest.vmresult == VMSUSPEND);
+  		okendpt(p->p_vmrequest.who, &proc_nr);
+		target = proc_addr(proc_nr);
+		p->p_vmrequest.vmresult = m_ptr->SVMCTL_VALUE;
+		vmassert(p->p_vmrequest.vmresult != VMSUSPEND);
+		if(p->p_vmrequest.vmresult != OK)
 			kprintf("SYSTEM: VM replied %d to mem request\n",
-				rp->p_vmrequest.vmresult);
+				p->p_vmrequest.vmresult);
 
+
+#if 0
+		printf("memreq reply: vm request sent by: %s / %d about %d; 0x%lx-0x%lx, wr %d, stack: %s ",
+			p->p_name, p->p_endpoint, p->p_vmrequest.who,
+			p->p_vmrequest.start,
+			p->p_vmrequest.start + p->p_vmrequest.length,
+			p->p_vmrequest.writeflag, p->p_vmrequest.stacktrace);
+		printf("type %d\n", p->p_vmrequest.type);
+#endif
+
+		vmassert(RTS_ISSET(target, VMREQTARGET));
+		RTS_LOCK_UNSET(target, VMREQTARGET);
+
+		if(p->p_vmrequest.type == VMSTYPE_KERNELCALL) {
 			/* Put on restart chain. */
-		rp->p_vmrequest.nextrestart = vmrestart;
-		vmrestart = rp;
-
+			p->p_vmrequest.nextrestart = vmrestart;
+			vmrestart = p;
+		} else if(p->p_vmrequest.type == VMSTYPE_DELIVERMSG) {
+			vmassert(p->p_misc_flags & MF_DELIVERMSG);
+			vmassert(p == target);
+			vmassert(RTS_ISSET(p, VMREQUEST));
+			vmassert(RTS_ISSET(p, VMREQTARGET));
+			RTS_LOCK_UNSET(p, VMREQUEST);
+			RTS_LOCK_UNSET(target, VMREQTARGET);
+		} else {
 #if DEBUG_VMASSERT
-		/* Sanity check. */
-		if(rp->p_vmrequest.vmresult == OK) {
-			if(CHECKRANGE(p,
-				rp->p_vmrequest.start,
-				rp->p_vmrequest.length,
-				rp->p_vmrequest.writeflag) != OK) {
-kprintf("SYSTEM: request %d:0x%lx-0x%lx, wrflag %d, failed\n",
-	rp->p_endpoint,
-	rp->p_vmrequest.start,  rp->p_vmrequest.start + rp->p_vmrequest.length,
-	rp->p_vmrequest.writeflag); 
+			printf("suspended with stack: %s\n",
+				p->p_vmrequest.stacktrace);
+#endif
+			minix_panic("strange request type",
+				p->p_vmrequest.type);
+		}
 
-				minix_panic("SYSTEM: fail but VM said OK", NO_NUM);
-			}
-		}
-#endif
 		return OK;
-#if VM_KERN_NOPAGEZERO
-	case VMCTL_NOPAGEZERO:
+	case VMCTL_ENABLE_PAGING:
+		if(vm_running) 
+			minix_panic("do_vmctl: paging already enabled", NO_NUM);
+		vm_init(p);
+		if(!vm_running)
+			minix_panic("do_vmctl: paging enabling failed", NO_NUM);
+		vmassert(p->p_delivermsg_lin ==
+		  umap_local(p, D, p->p_delivermsg_vir, sizeof(message)));
+		if(newmap(p, (struct mem_map *) m_ptr->SVMCTL_VALUE) != OK)
+			minix_panic("do_vmctl: newmap failed", NO_NUM);
+		FIXLINMSG(p);
+		vmassert(p->p_delivermsg_lin);
 		return OK;
-#endif
   }
 
   /* Try architecture-specific vmctls. */

kernel/table.c

@@ -35,7 +35,7 @@
 
 /* Define stack sizes for the kernel tasks included in the system image. */
 #define NO_STACK	0
-#define SMALL_STACK	(256 * sizeof(char *))
+#define SMALL_STACK	(1024 * sizeof(char *))
 #define IDL_S	SMALL_STACK	/* 3 intr, 3 temps, 4 db for Intel */
 #define	HRD_S	NO_STACK	/* dummy task, uses kernel stack */
 #define	TSK_S	SMALL_STACK	/* system and clock task */
@@ -96,7 +96,8 @@ PRIVATE int
   drv_c[] = { DRV_C },
   tty_c[] = { DRV_C, SYS_PHYSCOPY, SYS_ABORT, SYS_IOPENABLE,
 		SYS_READBIOS },
-  mem_c[] = { DRV_C, SYS_PHYSCOPY, SYS_PHYSVCOPY, SYS_IOPENABLE };
+  mem_c[] = { DRV_C, SYS_PHYSCOPY, SYS_PHYSVCOPY, SYS_IOPENABLE },
+  usr_c[] = { SYS_SYSCTL };
 
 /* The system image table lists all programs that are part of the boot image. 
  * The order of the entries here MUST agree with the order of the programs
@@ -117,16 +118,16 @@ PUBLIC struct boot_image image[] = {
 {CLOCK,clock_task,TSK_F,  8, TASK_Q, TSK_S, TSK_T,     0, no_c,"clock" },
 {SYSTEM, sys_task,TSK_F,  8, TASK_Q, TSK_S, TSK_T,     0, no_c,"system"},
 {HARDWARE,      0,TSK_F,  8, TASK_Q, HRD_S,     0,     0, no_c,"kernel"},
-{PM_PROC_NR,    0,SVM_F, 32,      4, 0,     SRV_T, SRV_M, c(pm_c),"pm"    },
-{FS_PROC_NR,    0,SVM_F, 32,      5, 0,     SRV_T, SRV_M, c(fs_c),"vfs"   },
+{PM_PROC_NR,    0,SRV_F, 32,      4, 0,     SRV_T, SRV_M, c(pm_c),"pm"    },
+{FS_PROC_NR,    0,SRV_F, 32,      5, 0,     SRV_T, SRV_M, c(fs_c),"vfs"   },
 {RS_PROC_NR,    0,SVM_F,  4,      4, 0,     SRV_T, SYS_M, c(rs_c),"rs"    },
-{DS_PROC_NR,    0,SVM_F,  4,      4, 0,     SRV_T, SYS_M, c(ds_c),"ds"    },
+{DS_PROC_NR,    0,SRV_F,  4,      4, 0,     SRV_T, SYS_M, c(ds_c),"ds"    },
 {TTY_PROC_NR,   0,SVM_F,  4,      1, 0,     SRV_T, SYS_M,c(tty_c),"tty"   },
 {MEM_PROC_NR,   0,SVM_F,  4,      3, 0,     SRV_T, SYS_M,c(mem_c),"memory"},
-{LOG_PROC_NR,   0,SVM_F,  4,      2, 0,     SRV_T, SYS_M,c(drv_c),"log"   },
+{LOG_PROC_NR,   0,SRV_F,  4,      2, 0,     SRV_T, SYS_M,c(drv_c),"log"   },
 {MFS_PROC_NR,   0,SVM_F, 32,      5, 0,     SRV_T, SRV_M, c(fs_c),"mfs"   },
 {VM_PROC_NR,    0,SRV_F, 32,      2, 0,     SRV_T, SRV_M, c(vm_c),"vm"    },
-{INIT_PROC_NR,  0,USR_F,  8, USER_Q, 0,     USR_T, USR_M, no_c,"init"  },
+{INIT_PROC_NR,  0,USR_F,  8, USER_Q, 0,     USR_T, USR_M, c(usr_c),"init"  },
 };
 
 /* Verify the size of the system image table at compile time. Also verify that 

kernel/utility.c

@@ -37,7 +37,7 @@ char *mess;
 int nr;
 {
 /* The system has run aground of a fatal kernel error. Terminate execution. */
-  if (minix_panicing ++) return;		/* prevent recursive panics */
+if (!minix_panicing++) {
 
   if (mess != NULL) {
 	kprintf("kernel panic: %s", mess);
@@ -46,8 +46,10 @@ int nr;
 	kprintf("\n");
   }
 
+  kprintf("proc_ptr %s / %d\n", proc_ptr->p_name, proc_ptr->p_endpoint);
   kprintf("kernel stacktrace: ");
   util_stacktrace();
+}
 
   /* Abort MINIX. */
   minix_shutdown(NULL);
@@ -80,8 +82,7 @@ int c;					/* character to append */
       kmess.km_next = (kmess.km_next + 1) % _KMESS_BUF_SIZE;
   } else {
       int p, outprocs[] = OUTPUT_PROCS_ARRAY;
-      if(do_serial_debug) return;
-      if(minix_panicing || do_serial_debug) return;
+      if(!(minix_panicing || do_serial_debug)) {
 	      for(p = 0; outprocs[p] != NONE; p++) {
 		 if(isokprocn(outprocs[p]) && !isemptyn(outprocs[p])) {
        	    send_sig(outprocs[p], SIGKMESS);
@@ -89,3 +90,5 @@ int c;					/* character to append */
       	}
      }
   }
+  return;
+}

kernel/vm.h

@@ -2,18 +2,19 @@
 #ifndef _VM_H
 #define _VM_H 1
 
-#define CHECKRANGE_OR_SUSPEND(pr, start, length, wr)  { int mr; \
-	if(vm_running && (mr=vm_checkrange(proc_addr(who_p), pr, start, length, wr, 0)) != OK) { \
-		return mr;					 \
-	} }
-
-#define CHECKRANGE(pr, start, length, wr)   \
-	vm_checkrange(proc_addr(who_p), pr, start, length, wr, 1)
-
-/* Pseudo error code indicating a process request has to be
- * restarted after an OK from VM.
- */
+/* Pseudo error codes */
 #define VMSUSPEND       -996
+#define EFAULT_SRC	-995
+#define EFAULT_DST	-994
+
+#define FIXLINMSG(prp) { prp->p_delivermsg_lin = umap_local(prp, D, prp->p_delivermsg_vir, sizeof(message)); }
+
+#define PHYS_COPY_CATCH(src, dst, size, a) {	\
+	vmassert(intr_disabled());		\
+	catch_pagefaults++;			\
+	a = phys_copy(src, dst, size);		\
+	catch_pagefaults--;			\
+	}
 
 #endif
 

lib/other/_lseek64.c

@@ -17,6 +17,6 @@ u64_t *newpos;
   m.m2_i2 = whence;
   if (_syscall(FS, LLSEEK, &m) < 0) return -1;
   if (newpos)
-	*newpos= make64(m.m2_l2, m.m2_l1);
+	*newpos= make64(m.m2_l1, m.m2_l2);
   return 0;
 }

lib/syslib/Makefile.in

@@ -66,7 +66,6 @@ libsys_FILES=" \
 	sys_vinl.c \
 	sys_vinw.c \
 	sys_vircopy.c \
-	sys_vm_setbuf.c \
 	sys_vmctl.c \
 	sys_voutb.c \
 	sys_voutl.c \

lib/syslib/sys_fork.c

@@ -1,11 +1,12 @@
 #include "syslib.h"
 
-PUBLIC int sys_fork(parent, child, child_endpoint, map_ptr, flags)
+PUBLIC int sys_fork(parent, child, child_endpoint, map_ptr, flags, msgaddr)
 endpoint_t parent;		/* process doing the fork */
 endpoint_t child;		/* which proc has been created by the fork */
 int *child_endpoint;
 struct mem_map *map_ptr;
 u32_t flags;
+vir_bytes *msgaddr;
 {
 /* A process has forked.  Tell the kernel. */
 
@@ -18,5 +19,6 @@ u32_t flags;
   m.PR_FORK_FLAGS = flags;
   r = _taskcall(SYSTASK, SYS_FORK, &m);
   *child_endpoint = m.PR_ENDPT;
+  *msgaddr = m.PR_FORK_MSGADDR;
   return r;
 }

lib/syslib/sys_vm_setbuf.c

@@ -1,21 +0,0 @@
-#include "syslib.h"
-
-/*===========================================================================*
- *                                sys_vm_setbuf				     *
- *===========================================================================*/
-PUBLIC int sys_vm_setbuf(base, size, high)
-phys_bytes base;
-phys_bytes size;
-phys_bytes high;
-{
-    message m;
-    int result;
-
-    m.m4_l1= base;
-    m.m4_l2= size;
-    m.m4_l3= high;
-
-    result = _taskcall(SYSTASK, SYS_VM_SETBUF, &m);
-    return(result);
-}
-

lib/syslib/sys_vmctl.c

@@ -43,7 +43,7 @@ PUBLIC int sys_vmctl_get_cr3_i386(endpoint_t who, u32_t *cr3)
 }
 
 PUBLIC int sys_vmctl_get_memreq(endpoint_t *who, vir_bytes *mem,
-	vir_bytes *len, int *wrflag)
+	vir_bytes *len, int *wrflag, endpoint_t *requestor)
 {
   message m;
   int r;
@@ -56,7 +56,16 @@ PUBLIC int sys_vmctl_get_memreq(endpoint_t *who, vir_bytes *mem,
 	*mem = (vir_bytes) m.SVMCTL_MRG_ADDR;
 	*len = m.SVMCTL_MRG_LEN;
 	*wrflag = m.SVMCTL_MRG_WRITE;
+	*requestor = (endpoint_t) m.SVMCTL_MRG_REQUESTOR;
   }
   return r;
 }
 
+PUBLIC int sys_vmctl_enable_paging(struct mem_map *map)
+{
+	message m;
+	m.SVMCTL_WHO = SELF;
+	m.SVMCTL_PARAM = VMCTL_ENABLE_PAGING;
+	m.SVMCTL_VALUE = (int) map;
+	return _taskcall(SYSTASK, SYS_VMCTL, &m);
+}

lib/sysutil/Makefile.in

@@ -6,7 +6,6 @@ LIBRARIES=libsys
 
 libsys_FILES=" \
 	asynsend.c \
-	kmalloc.c \
 	kprintf.c \
 	kputc.c \
 	tickdelay.c \

lib/sysutil/kmalloc.c

@@ -1,177 +0,0 @@
-/*	malloc(), realloc(), free() - simple memory allocation routines
- *
- * This is a very small and simple minded malloc	Author: Kees J. Bot
- * implementation.  Ideal for things like a			29 Jan 1994
- * bootstrap program, or for debugging.  Six times
- * slower than any good malloc.
- */
-#define nil 0
-
-#define sbrk _sbrk
-#include <stddef.h>
-#include <stdlib.h>
-#include <unistd.h>
-#include <string.h>
-#include <limits.h>
-#if !DEBUG
-#define NDEBUG	1
-#define debug(expr)	((void) 0)
-#else
-#define debug(expr)	expr
-#endif
-#include <assert.h>
-
-typedef struct cell {
-	size_t		size;		/* Size of a malloc()'ed object. */
-#if DEBUG
-	unsigned	magic;		/* To recognize a cell. */
-#endif
-	struct cell	*next;		/* Next cell on the free list. */
-#if DEBUG
-	unsigned	sacred;		/* Don't touch while unallocated. */
-#endif
-} cell_t;
-
-#if UINT_MAX <= 0xFFFF
-#define MAGIC	0x537B
-#else
-#define MAGIC	0x537BC0D8
-#endif
-
-/* Size of the header of an object. */
-#define HDR_SIZE	offsetof(cell_t, next)
-
-/* An offset from a cell pointer to a next cell. */
-#define offset(cp, size)	((cell_t *) ((char *) (cp) + (size)))
-
-/* Address of the object in a cell and back. */
-#define cell2obj(cp)		((void *) ((char *) (cp) + HDR_SIZE))
-#define obj2cell(op)		((cell_t *) ((char *) (op) - HDR_SIZE))
-
-/* The free list. */
-static cell_t *freelist;
-
-void *malloc(size_t size)
-/* Allocate an object of at least the given size. */
-{
-	cell_t **pcp, *cp;
-
-	size += HDR_SIZE;
-	if (size < sizeof(cell_t)) size= sizeof(cell_t);
-
-	/* Align to a word.  Use a real malloc if you need better alignment. */
-	size= (size + sizeof(int) - 1) & ~(sizeof(int) - 1);
-
-	/* Space for a magic number at the end of the chunk. */
-	debug(size += sizeof(unsigned));
-
-	for (;;) {
-		/* Do a first fit search. */
-		pcp= &freelist;
-		while ((cp= *pcp) != nil) {
-			cell_t *next= cp->next;
-
-			assert(cp->magic == MAGIC);
-			assert(cp->sacred == MAGIC);
-
-			if (offset(cp, cp->size) == next) {
-				/* Join adjacent free cells. */
-				assert(next->magic == MAGIC);
-				assert(next->sacred == MAGIC);
-
-				cp->size+= next->size;
-				cp->next= next->next;
-
-				continue;		/* Try again. */
-			}
-			if (size <= cp->size) break;	/* Big enough. */
-
-			/* Next cell. */
-			pcp= &cp->next;
-		}
-
-		if (cp != nil) break;	/* Found a big enough chunk. */
-
-		/* Allocate a new chunk at the break. */
-		if ((cp= (cell_t *) sbrk(size)) == (cell_t *) -1) {
-			return nil;
-		}
-
-		cp->size= size;
-		cp->next= nil;
-		debug(cp->magic= MAGIC);
-		debug(cp->sacred= MAGIC);
-		*pcp= cp;
-	}
-
-	/* We've got a cell that is big enough.  Can we break it up? */
-	if (cp->size >= size + sizeof(cell_t)) {
-		cell_t *next= offset(cp, size);
-
-		next->size= cp->size - size;
-		next->next= cp->next;
-		debug(next->magic= MAGIC);
-		debug(next->sacred= MAGIC);
-		cp->size= size;
-		cp->next= next;
-	}
-
-	/* Unchain the cell we've found and return an address in it. */
-	*pcp= cp->next;
-	debug(memset(cell2obj(cp), 0xAA, cp->size - HDR_SIZE));
-	debug(((unsigned *) offset(cp, cp->size))[-1]= MAGIC);
-
-	return cell2obj(cp);
-}
-
-void free(void *op)
-/* Deallocate an object. */
-{
-	cell_t **prev, *next, *cp;
-
-	if (op == nil) return;		/* Aaargh. */
-
-	cp= obj2cell(op);
-	assert(cp->magic == MAGIC);
-	assert(((unsigned *) offset(cp, cp->size))[-1] == MAGIC);
-	debug(cp->sacred= MAGIC);
-
-	/* Find the spot where the object belongs. */
-	prev= &freelist;
-	while ((next= *prev) != nil && next < cp) {
-		assert(next->magic == MAGIC);
-		assert(next->sacred == MAGIC);
-		prev= &next->next;
-	}
-
-	/* Put the new free cell in the list. */
-	*prev= cp;
-	cp->next= next;
-
-#if DEBUG
-	/* Check the rest of the list. */
-	while (next != nil) {
-		assert(next->magic == MAGIC);
-		assert(next->sacred == MAGIC);
-		next= next->next;
-	}
-#endif
-}
-
-void *realloc(void *op, size_t size)
-/* Change the size of an object.  Don't bother being smart, just copy it. */
-{
-	size_t oldsize;
-	void *new;
-
-	oldsize= op == nil ? 0 : obj2cell(op)->size - HDR_SIZE;
-
-	new= malloc(size);
-	memcpy(new, op, oldsize > size ? size : oldsize);
-	free(op);
-	return new;
-}
-
-/*
- * $PchId: malloc.c,v 1.4 1996/02/22 09:15:56 philip Exp $
- */

man/man1/ls.1

@@ -2,7 +2,7 @@
 .SH NAME
 ls \- list the contents of a directory
 .SH SYNOPSIS
-\fBls\fP [\fB\-acdfghilpqrstu1ACDFLMRTX\fP] [\fIname\fP...]
+\fBls\fP [\fB\-acdfghilnpqrstu1ACDFLMRTX\fP] [\fIname\fP...]
 .SH DESCRIPTION
 For each file argument, list it.  For each directory argument, list its
 contents.  The current working directory is listed when no files are named.
@@ -30,7 +30,7 @@ and
 .PP
 Files whose names start with a dot are by default not listed.
 .PP
-Note that standard MINIX 3 doesn't have symbolic links or sockets and
+Note that standard MINIX 3 doesn't have sockets, and
 .B \-u
 and
 .B \-c

servers/is/dmp.c

@@ -21,7 +21,6 @@ struct hook_entry {
 	{ F5,	monparams_dmp, "Boot monitor parameters" },
 	{ F6,	irqtab_dmp, "IRQ hooks and policies" },
 	{ F7,	kmessages_dmp, "Kernel messages" },
-	{ F9,	sched_dmp, "Scheduling queues" },
 	{ F10,	kenv_dmp, "Kernel parameters" },
 	{ F11,	timing_dmp, "Timing details (if enabled)" },
 	{ SF1,	mproc_dmp, "Process manager process table" },

servers/is/dmp_kernel.c

@@ -224,54 +224,6 @@ PUBLIC void image_dmp()
   printf("\n");
 }
 
-/*===========================================================================*
- *				sched_dmp    				     *
- *===========================================================================*/
-PUBLIC void sched_dmp()
-{
-  struct proc *rdy_head[NR_SCHED_QUEUES];
-  struct kinfo kinfo;
-  register struct proc *rp;
-  vir_bytes ptr_diff;
-  int r;
-
-  /* First obtain a scheduling information. */
-  if ((r = sys_getschedinfo(proc, rdy_head)) != OK) {
-      report("IS","warning: couldn't get copy of process table", r);
-      return;
-  }
-  /* Then obtain kernel addresses to correct pointer information. */
-  if ((r = sys_getkinfo(&kinfo)) != OK) {
-      report("IS","warning: couldn't get kernel addresses", r);
-      return;
-  }
-
-  /* Update all pointers. Nasty pointer algorithmic ... */
-  ptr_diff = (vir_bytes) proc - (vir_bytes) kinfo.proc_addr;
-  for (r=0;r<NR_SCHED_QUEUES; r++)
-      if (rdy_head[r] != NIL_PROC)
-          rdy_head[r] = 
-              (struct proc *)((vir_bytes) rdy_head[r] + ptr_diff);
-  for (rp=BEG_PROC_ADDR; rp < END_PROC_ADDR; rp++)
-      if (rp->p_nextready != NIL_PROC)
-          rp->p_nextready =
-               (struct proc *)((vir_bytes) rp->p_nextready + ptr_diff);
-
-  /* Now show scheduling queues. */
-  printf("Dumping scheduling queues.\n");
-
-  for (r=0;r<NR_SCHED_QUEUES; r++) {
-      rp = rdy_head[r];
-      if (!rp) continue;
-      printf("%2d: ", r);
-      while (rp != NIL_PROC) {
-          printf("%3d ", rp->p_nr);
-          rp = rp->p_nextready;
-      }
-      printf("\n");
-  }
-  printf("\n");
-}
 
 /*===========================================================================*
  *				kenv_dmp				     *
@@ -311,9 +263,6 @@ PUBLIC void kenv_dmp()
     printf("- nr_tasks:     %3u\n", kinfo.nr_tasks); 
     printf("- release:      %.6s\n", kinfo.release); 
     printf("- version:      %.6s\n", kinfo.version); 
-#if DEBUG_LOCK_CHECK
-    printf("- relocking:    %d\n", kinfo.relocking); 
-#endif
     printf("\n");
 }
 

servers/mfs/Makefile

@@ -25,7 +25,6 @@ OBJ =	cache.o device.o link.o \
 all build:	$(SERVER)
 $(SERVER):	$(OBJ)
 	$(CC) -o $@ $(LDFLAGS) $(OBJ) $(LIBS)
-	install -S `expr $(NR_BUFS) \* $(BS) \* 2.2` $(SERVER)
 
 install: $(SERVER)
 	-mv $(DEST) $(DEST).prev

servers/mfs/cache.c

@@ -17,6 +17,7 @@
 #include "fs.h"
 #include <minix/com.h>
 #include <minix/u64.h>
+#include <string.h>
 #include "buf.h"
 #include "super.h"
 #include "inode.h"
@@ -24,6 +25,9 @@
 FORWARD _PROTOTYPE( void rm_lru, (struct buf *bp) );
 FORWARD _PROTOTYPE( int rw_block, (struct buf *, int) );
 
+char saved[100];
+char *savedptr;
+
 /*===========================================================================*
  *				get_block				     *
  *===========================================================================*/
@@ -48,10 +52,19 @@ int only_search;		/* if NO_READ, don't read, else act normal */
  */
 
   int b;
-  register struct buf *bp, *prev_ptr;
+  static struct buf *bp, *prev_ptr;
+
+  savedptr = (vir_bytes) &b + sizeof(b);
+  memcpy(saved, savedptr, sizeof(saved));
+
+#define CHECK \
+  if(memcmp(saved, savedptr, sizeof(saved))) \
+	panic(__FILE__,"memory corruption", __LINE__); 
 
   ASSERT(fs_block_size > 0);
 
+  CHECK;
+
   /* Search the hash chain for (dev, block). Do_read() can use 
    * get_block(NO_DEV ...) to get an unnamed block to fill with zeros when
    * someone wants to read from a hole in a file, in which case this search
@@ -60,8 +73,11 @@ int only_search;		/* if NO_READ, don't read, else act normal */
   if (dev != NO_DEV) {
 	b = BUFHASH(block);
 	bp = buf_hash[b];
+  CHECK;
 	while (bp != NIL_BUF) {
+  CHECK;
 		if (bp->b_blocknr == block && bp->b_dev == dev) {
+  CHECK;
 			/* Block needed has been found. */
 			if (bp->b_count == 0) rm_lru(bp);
 			bp->b_count++;	/* record that block is in use */
@@ -69,6 +85,7 @@ int only_search;		/* if NO_READ, don't read, else act normal */
 			ASSERT(bp->b_dev == dev);
 			ASSERT(bp->b_dev != NO_DEV);
 			ASSERT(bp->bp);
+  CHECK;
 			return(bp);
 		} else {
 			/* This block is not the one sought. */
@@ -137,6 +154,7 @@ int only_search;		/* if NO_READ, don't read, else act normal */
 
   ASSERT(bp->bp);
 
+  CHECK;
   return(bp);			/* return the newly acquired block */
 }
 

servers/pm/misc.c

@@ -346,12 +346,9 @@ PUBLIC int do_getprocnr()
 		if (((rmp->mp_flags & (IN_USE | ZOMBIE)) == IN_USE) && 
 			strncmp(rmp->mp_name, search_key, key_len)==0) {
   			mp->mp_reply.endpt = rmp->mp_endpoint;
-  			printf("PM: name %s result: %d\n", search_key, 
-					rmp->mp_endpoint);
   			return(OK);
 		} 
 	}
-	printf("PM: name %s result: ESRCH\n", search_key);
   	return(ESRCH);			
   } else {			/* return own/parent process number */
 #if 0

servers/pm/signal.c

@@ -528,7 +528,8 @@ doterminate:
 
   rmp->mp_sigstatus = (char) signo;
   if (sigismember(&core_sset, signo) && slot != FS_PROC_NR) {
-	printf("PM: signal %d for %d / %s\n", signo, rmp->mp_pid, rmp->mp_name);
+	printf("PM: signal %d for pid %d / %s\n",
+		signo, rmp->mp_pid, rmp->mp_name);
 	s= dump_core(rmp);
 	if (s == SUSPEND) {
 		return;

servers/vfs/main.c

@@ -26,6 +26,7 @@
 #include <minix/const.h>
 #include <minix/endpoint.h>
 #include <minix/safecopies.h>
+#include <minix/debug.h>
 #include "file.h"
 #include "fproc.h"
 #include "param.h"
@@ -59,6 +60,10 @@ PUBLIC int main()
 
   SANITYCHECK;
 
+#if DO_SANITYCHECKS
+  FIXME("VFS: DO_SANITYCHECKS is on");
+#endif
+
   /* This is the main loop that gets work, processes it, and sends replies. */
   while (TRUE) {
 	SANITYCHECK;
@@ -285,9 +290,14 @@ PRIVATE void get_work()
 	continue;
     }
     if(who_p >= 0 && fproc[who_p].fp_endpoint != who_e) {
-    	printf("FS: receive endpoint inconsistent (%d, %d, %d).\n",
-		who_e, fproc[who_p].fp_endpoint, who_e);
+	if(fproc[who_p].fp_endpoint == NONE) { 
+		printf("slot unknown even\n");
+	}
+    	printf("FS: receive endpoint inconsistent (source %d, who_p %d, stored ep %d, who_e %d).\n",
+		m_in.m_source, who_p, fproc[who_p].fp_endpoint, who_e);
+#if 0
 	panic(__FILE__, "FS: inconsistent endpoint ", NO_NUM);
+#endif
 	continue;
     }
     call_nr = m_in.m_type;

servers/vm/Makefile

@@ -4,7 +4,7 @@ SERVER = vm
 include /etc/make.conf
 
 OBJ = main.o alloc.o utility.o exec.o exit.o fork.o break.o \
-	signal.o vfs.o mmap.o slaballoc.o region.o pagefaults.o
+	signal.o vfs.o mmap.o slaballoc.o region.o pagefaults.o addravl.o 
 ARCHOBJ =  $(ARCH)/vm.o $(ARCH)/pagetable.o $(ARCH)/arch_pagefaults.o $(ARCH)/util.o 
 
 CPPFLAGS=-I../../kernel/arch/$(ARCH)/include -I$(ARCH)

servers/vm/addravl.c

@@ -0,0 +1,5 @@
+
+#include "pagerange.h"
+#include "addravl.h"
+#include "cavl_impl.h"
+

servers/vm/addravl.h

@@ -0,0 +1,17 @@
+
+#define AVL_UNIQUE(id) addr_ ## id
+#define AVL_HANDLE pagerange_t *
+#define AVL_KEY phys_bytes
+#define AVL_MAX_DEPTH 30 /* good for 2 million nodes */
+#define AVL_NULL NULL
+#define AVL_GET_LESS(h, a) (h)->less
+#define AVL_GET_GREATER(h, a) (h)->greater
+#define AVL_SET_LESS(h1, h2) (h1)->less = h2;
+#define AVL_SET_GREATER(h1, h2) (h1)->greater = h2;
+#define AVL_GET_BALANCE_FACTOR(h) (h)->factor
+#define AVL_SET_BALANCE_FACTOR(h, f) (h)->factor = f;
+#define AVL_COMPARE_KEY_KEY(k1, k2) ((k1) > (k2) ? 1 : ((k1) < (k2) ? -1 : 0))
+#define AVL_COMPARE_KEY_NODE(k, h) AVL_COMPARE_KEY_KEY((k), (h)->addr)
+#define AVL_COMPARE_NODE_NODE(h1, h2) AVL_COMPARE_KEY_KEY((h1)->addr, (h2)->addr)
+
+#include "cavl_if.h"

servers/vm/alloc.c

@@ -23,6 +23,7 @@
 #include <minix/const.h>
 #include <minix/sysutil.h>
 #include <minix/syslib.h>
+#include <minix/debug.h>
 
 #include <sys/mman.h>
 
@@ -36,9 +37,11 @@
 #include "proto.h"
 #include "util.h"
 #include "glo.h"
+#include "pagerange.h"
+#include "addravl.h"
 
-/* Initially, no free pages are known. */
-PRIVATE phys_bytes free_pages_head = NO_MEM;	/* Physical address in bytes. */
+/* AVL tree of free pages. */
+addr_avl addravl;
 
 /* Used for sanity check. */
 PRIVATE phys_bytes mem_low, mem_high;
@@ -102,26 +105,6 @@ FORWARD _PROTOTYPE( void holes_sanity_f, (char *fn, int line)		    );
 }
 
 
-void availbytes(vir_bytes *bytes, vir_bytes *chunks)
-{
-	phys_bytes p, nextp;
-	*bytes = 0;
-	*chunks = 0;
-	for(p = free_pages_head; p != NO_MEM; p = nextp) {
-		phys_bytes thissize, ret;
-		GET_PARAMS(p, thissize, nextp);
-		(*bytes) += thissize;
-		(*chunks)++;
-		if(nextp != NO_MEM)	{
-			vm_assert(nextp > p);
-			vm_assert(nextp > p + thissize);
-		}
-	}
-
-	return;
-}
-
-
 #if SANITYCHECKS
 
 /*===========================================================================*
@@ -410,6 +393,8 @@ struct memory *chunks;		/* list of free memory chunks */
   hole_head = NIL_HOLE;
   free_slots = &hole[0];
 
+  addr_init(&addravl);
+
   /* Use the chunks of physical memory to allocate holes. */
   for (i=NR_MEMS-1; i>=0; i--) {
   	if (chunks[i].size > 0) {
@@ -425,213 +410,125 @@ struct memory *chunks;		/* list of free memory chunks */
   CHECKHOLES;
 }
 
+int countnodes(void)
+{
+	addr_iter iter;
+	int n = 0;
+	addr_start_iter_least(&addravl, &iter);
+	while(addr_get_iter(&iter)) {
+		n++;
+		addr_incr_iter(&iter);
+	}
+
+	return n;
+}
+
 /*===========================================================================*
  *				alloc_pages				     *
  *===========================================================================*/
 PRIVATE PUBLIC phys_bytes alloc_pages(int pages, int memflags)
 {
-	phys_bytes bytes, p, nextp, prevp = NO_MEM;
-	phys_bytes prevsize = 0;
+	addr_iter iter;
+	pagerange_t *pr;
+	int incr;
+	phys_bytes boundary16 = 16 * 1024 * 1024 / VM_PAGE_SIZE;
+	phys_bytes mem;
 
-#if SANITYCHECKS
-	vir_bytes avail1, avail2, chunks1, chunks2;
-	availbytes(&avail1, &chunks1);
-#endif
-
-	vm_assert(pages > 0);
-	bytes = CLICK2ABS(pages);
-	vm_assert(ABS2CLICK(bytes) == pages);
-
-#if SANITYCHECKS
-#define ALLOCRETURNCHECK			\
-	availbytes(&avail2, &chunks2);		\
-	vm_assert(avail1 - bytes == avail2);	\
-	vm_assert(chunks1 == chunks2 || chunks1-1 == chunks2);
-#else
-#define ALLOCRETURNCHECK
-#endif
-
-
-	for(p = free_pages_head; p != NO_MEM; p = nextp) {
-		phys_bytes thissize, ret;
-		GET_PARAMS(p, thissize, nextp);
-		if(thissize >= bytes) {
-			/* We found a chunk that's big enough. */
-
-			ret = p + thissize - bytes;
-			thissize -= bytes;
-
-			if(thissize == 0) {
-				/* Special case: remove this link entirely. */
-				if(prevp == NO_MEM)
-					free_pages_head = nextp;
-				else {
-					vm_assert(prevsize > 0);
-					SET_PARAMS(prevp, prevsize, nextp);
-				}
+	if(memflags & PAF_LOWER16MB) {
+		addr_start_iter_least(&addravl, &iter);
+		incr = 1;
 	} else {
-				/* Remove memory from this chunk. */
-				SET_PARAMS(p, thissize, nextp);
+		addr_start_iter_greatest(&addravl, &iter);
+		incr = 0;
+	}
+
+	while((pr = addr_get_iter(&iter))) {
+		SLABSANE(pr);
+		if(pr->size >= pages) {
+			if(memflags & PAF_LOWER16MB) {
+				if(pr->addr + pages > boundary16)
+					return NO_MEM;
+			}
+
+			/* good block found! */
+			break;
+		}
+		if(incr)
+			addr_incr_iter(&iter);
+		else
+			addr_decr_iter(&iter);
+	}
+
+	if(!pr)
+		return NO_MEM;
+
+	SLABSANE(pr);
+
+	mem = pr->addr;
+
+	vm_assert(pr->size >= pages);
+	if(pr->size == pages) {
+		pagerange_t *prr;
+		prr = addr_remove(&addravl, pr->addr);
+		vm_assert(prr);
+		vm_assert(prr == pr);
+		SLABFREE(pr);
+	} else {
+		pr->addr += pages;
+		pr->size -= pages;
 	}
 
-			/* Clear memory if requested. */
 	if(memflags & PAF_CLEAR) {
 		int s;
-			  if ((s= sys_memset(0, ret, bytes)) != OK)   {
-				vm_panic("alloc_pages: sys_memset failed", s);
-			  }
+		if ((s= sys_memset(0, CLICK_SIZE*mem,
+			VM_PAGE_SIZE*pages)) != OK) 
+			vm_panic("alloc_mem: sys_memset failed", s);
 	}
 
-			/* Check if returned range is actual good memory. */
-			vm_assert_range(ret, bytes);
-
-			ALLOCRETURNCHECK;
-
-			/* Return it in clicks. */
-			return ABS2CLICK(ret);
-		}
-		prevp = p;
-		prevsize = thissize;
-	}
-	return NO_MEM;
+	return mem;
 }
 
 /*===========================================================================*
  *				free_pages				     *
  *===========================================================================*/
-PRIVATE PUBLIC void free_pages(phys_bytes pageno, int npages)
+PRIVATE void free_pages(phys_bytes pageno, int npages)
 {
-	phys_bytes p, origsize,
-		size, nextaddr, thissize, prevp = NO_MEM, pageaddr;
+	pagerange_t *pr, *p;
+	addr_iter iter;
 
-#if SANITYCHECKS
-	vir_bytes avail1, avail2, chunks1, chunks2;
-	availbytes(&avail1, &chunks1);
-#endif
+	vm_assert(!addr_search(&addravl, pageno, AVL_EQUAL));
 
-#if SANITYCHECKS
-#define FREERETURNCHECK								\
-	availbytes(&avail2, &chunks2);					\
-	vm_assert(avail1 + origsize  == avail2);			\
-	vm_assert(chunks1 == chunks2 || chunks1+1 == chunks2 || chunks1-1 == chunks2);
-#else
-#define FREERETURNCHECK
-#endif
-
-	/* Basic sanity check. */
+	/* try to merge with higher neighbour */
+	if((pr=addr_search(&addravl, pageno+npages, AVL_EQUAL))) {
+		SLABSANE(pr);
+		pr->addr -= npages;
+		pr->size += npages;
+	} else {
+		if(!SLABALLOC(pr))
+			vm_panic("alloc_pages: can't alloc", NO_NUM);
+		SLABSANE(pr);
 		vm_assert(npages > 0);
-	vm_assert(pageno != NO_MEM);	/* Page number must be reasonable. */
-
-	/* Convert page and pages to bytes. */
-	pageaddr = CLICK2ABS(pageno);
-	origsize = size = npages * VM_PAGE_SIZE;	/* Size in bytes. */
-	vm_assert(pageaddr != NO_MEM);
-	vm_assert(ABS2CLICK(pageaddr) == pageno);
-	vm_assert_range(pageaddr, size);
-
-	/* More sanity checks. */
-	vm_assert(ABS2CLICK(size) == npages);	/* Sanity. */
-	vm_assert(pageaddr + size > pageaddr);		/* Must not overflow. */
-
-	/* Special case: no free pages. */
-	if(free_pages_head == NO_MEM) {
-		free_pages_head = pageaddr;
-		SET_PARAMS(pageaddr, size, NO_MEM);
-		FREERETURNCHECK;
-		return;
+		pr->addr = pageno;
+		pr->size = npages;
+		addr_insert(&addravl, pr);
 	}
 
-	/* Special case: the free block is before the current head. */
-	if(pageaddr < free_pages_head) {
-		phys_bytes newsize, newnext, headsize, headnext;
-		vm_assert(pageaddr + size <= free_pages_head);
-		GET_PARAMS(free_pages_head, headsize, headnext);
-		newsize = size;
-		if(pageaddr + size == free_pages_head) {
-			/* Special case: contiguous. */
-			newsize += headsize;
-			newnext = headnext;
-		} else {
-			newnext = free_pages_head;
+	addr_start_iter(&addravl, &iter, pr->addr, AVL_EQUAL);
+	p = addr_get_iter(&iter);
+	vm_assert(p);
+	vm_assert(p == pr);
+
+	addr_decr_iter(&iter);
+	if((p = addr_get_iter(&iter))) {
+		SLABSANE(p);
+		if(p->addr + p->size == pr->addr) {
+			p->size += pr->size;
+			addr_remove(&addravl, pr->addr);
+			SLABFREE(pr);
 		}
-		SET_PARAMS(pageaddr, newsize, newnext);
-		free_pages_head = pageaddr;
-		FREERETURNCHECK;
-		return;
 	}
 
-	/* Find where to put the block in the free list. */
-	for(p = free_pages_head; p < pageaddr; p = nextaddr) {
-		GET_PARAMS(p, thissize, nextaddr);
-
-		if(nextaddr == NO_MEM) {
-			/* Special case: page is at the end of the list. */
-			if(p + thissize == pageaddr) {
-				/* Special case: contiguous. */
-				SET_PARAMS(p, thissize + size, NO_MEM);
-				FREERETURNCHECK;
-			} else {
-				SET_PARAMS(p, thissize, pageaddr);
-				SET_PARAMS(pageaddr, size, NO_MEM);
-				FREERETURNCHECK;
 }
-			return;
-		}
-
-		prevp = p;
-	}
-
-	/* Normal case: insert page block between two others.
-	 * The first block starts at 'prevp' and is 'thissize'.
-	 * The second block starts at 'p' and is 'nextsize'.
-	 * The block that has to come in between starts at
-	 * 'pageaddr' and is size 'size'.
-	 */
-	vm_assert(p != NO_MEM);
-	vm_assert(prevp != NO_MEM);
-	vm_assert(prevp < p);
-	vm_assert(p == nextaddr);
-
-#if SANITYCHECKS
-  {
-	vir_bytes prevpsize, prevpnext;
-	GET_PARAMS(prevp, prevpsize, prevpnext);
-	vm_assert(prevpsize == thissize);
-	vm_assert(prevpnext == p);
-
-	availbytes(&avail2, &chunks2);
-	vm_assert(avail1 == avail2);
-  }
-#endif
-
-	if(prevp + thissize == pageaddr) {
-		/* Special case: first block is contiguous with freed one. */
-		phys_bytes newsize = thissize + size;
-		SET_PARAMS(prevp, newsize, p);
-		pageaddr = prevp;
-		size = newsize;
-	} else {
-		SET_PARAMS(prevp, thissize, pageaddr);
-	}
-
-	/* The block has been inserted (and possibly merged with the
-	 * first one). Check if it has to be merged with the second one.
-	 */
-
-	if(pageaddr + size == p) {
-		phys_bytes nextsize, nextnextaddr;
-		/* Special case: freed block is contiguous with next one. */
-		GET_PARAMS(p, nextsize, nextnextaddr);
-		SET_PARAMS(pageaddr, size+nextsize, nextnextaddr);
-		FREERETURNCHECK;
-	} else {
-		SET_PARAMS(pageaddr, size, p);
-		FREERETURNCHECK;
-	}
-
-	return;
-}
-
 
 #define NR_DMA	16
 

servers/vm/cavl_if.h

@@ -0,0 +1,216 @@
+/* Abstract AVL Tree Generic C Package.
+** Interface generation header file.
+**
+** This code is in the public domain.  See cavl_tree.html for interface
+** documentation.
+**
+** Version: 1.5  Author: Walt Karas
+*/
+
+/* This header contains the definition of CHAR_BIT (number of bits in a
+** char). */
+#include <limits.h>
+
+#undef L__
+#undef L__EST_LONG_BIT
+#undef L__SIZE
+#undef L__SC
+#undef L__LONG_BIT
+#undef L__BIT_ARR_DEFN
+
+#ifndef AVL_SEARCH_TYPE_DEFINED_
+#define AVL_SEARCH_TYPE_DEFINED_
+
+typedef enum
+  {
+    AVL_EQUAL = 1,
+    AVL_LESS = 2,
+    AVL_GREATER = 4,
+    AVL_LESS_EQUAL = AVL_EQUAL | AVL_LESS,
+    AVL_GREATER_EQUAL = AVL_EQUAL | AVL_GREATER
+  }
+avl_search_type;
+
+#endif
+
+#ifdef AVL_UNIQUE
+
+#define L__ AVL_UNIQUE
+
+#else
+
+#define L__(X) X
+
+#endif
+
+/* Determine storage class for function prototypes. */
+#ifdef AVL_PRIVATE
+
+#define L__SC static
+
+#else
+
+#define L__SC extern
+
+#endif
+
+#ifdef AVL_SIZE
+
+#define L__SIZE AVL_SIZE
+
+#else
+
+#define L__SIZE unsigned long
+
+#endif
+
+typedef struct
+  {
+    #ifdef AVL_INSIDE_STRUCT
+
+    AVL_INSIDE_STRUCT
+
+    #endif
+
+    AVL_HANDLE root;
+  }
+L__(avl);
+
+/* Function prototypes. */
+
+L__SC void L__(init)(L__(avl) *tree);
+
+L__SC int L__(is_empty)(L__(avl) *tree);
+
+L__SC AVL_HANDLE L__(insert)(L__(avl) *tree, AVL_HANDLE h);
+
+L__SC AVL_HANDLE L__(search)(L__(avl) *tree, AVL_KEY k, avl_search_type st);
+
+L__SC AVL_HANDLE L__(search_least)(L__(avl) *tree);
+
+L__SC AVL_HANDLE L__(search_greatest)(L__(avl) *tree);
+
+L__SC AVL_HANDLE L__(remove)(L__(avl) *tree, AVL_KEY k);
+
+L__SC AVL_HANDLE L__(subst)(L__(avl) *tree, AVL_HANDLE new_node);
+
+#ifdef AVL_BUILD_ITER_TYPE
+
+L__SC int L__(build)(
+  L__(avl) *tree, AVL_BUILD_ITER_TYPE p, L__SIZE num_nodes);
+
+#endif
+
+/* ANSI C/ISO C++ require that a long have at least 32 bits.  Set
+** L__EST_LONG_BIT to be the greatest multiple of 8 in the range
+** 32 - 64 (inclusive) that is less than or equal to the number of
+** bits in a long.
+*/
+
+#if (((LONG_MAX >> 31) >> 7) == 0)
+
+#define L__EST_LONG_BIT 32
+
+#elif (((LONG_MAX >> 31) >> 15) == 0)
+
+#define L__EST_LONG_BIT 40
+
+#elif (((LONG_MAX >> 31) >> 23) == 0)
+
+#define L__EST_LONG_BIT 48
+
+#elif (((LONG_MAX >> 31) >> 31) == 0)
+
+#define L__EST_LONG_BIT 56
+
+#else
+
+#define L__EST_LONG_BIT 64
+
+#endif
+
+/* Number of bits in a long. */
+#define L__LONG_BIT (sizeof(long) * CHAR_BIT)
+
+/* The macro L__BIT_ARR_DEFN defines a bit array whose index is a (0-based)
+** node depth.  The definition depends on whether the maximum depth is more
+** or less than the number of bits in a single long.
+*/
+
+#if ((AVL_MAX_DEPTH) > L__EST_LONG_BIT)
+
+/* Maximum depth may be more than number of bits in a long. */
+
+#define L__BIT_ARR_DEFN(NAME) \
+  unsigned long NAME[((AVL_MAX_DEPTH) + L__LONG_BIT - 1) / L__LONG_BIT];
+
+#else
+
+/* Maximum depth is definitely less than number of bits in a long. */
+
+#define L__BIT_ARR_DEFN(NAME) unsigned long NAME;
+
+#endif
+
+/* Iterator structure. */
+typedef struct
+  {
+    /* Tree being iterated over. */
+    L__(avl) *tree_;
+
+    /* Records a path into the tree.  If bit n is true, indicates
+    ** take greater branch from the nth node in the path, otherwise
+    ** take the less branch.  bit 0 gives branch from root, and
+    ** so on. */
+    L__BIT_ARR_DEFN(branch)
+
+    /* Zero-based depth of path into tree. */
+    unsigned depth;
+
+    /* Handles of nodes in path from root to current node (returned by *). */
+    AVL_HANDLE path_h[(AVL_MAX_DEPTH) - 1];
+  }
+L__(iter);
+
+/* Iterator function prototypes. */
+
+L__SC void L__(start_iter)(
+  L__(avl) *tree, L__(iter) *iter, AVL_KEY k, avl_search_type st);
+
+L__SC void L__(start_iter_least)(L__(avl) *tree, L__(iter) *iter);
+
+L__SC void L__(start_iter_greatest)(L__(avl) *tree, L__(iter) *iter);
+
+L__SC AVL_HANDLE L__(get_iter)(L__(iter) *iter);
+
+L__SC void L__(incr_iter)(L__(iter) *iter);
+
+L__SC void L__(decr_iter)(L__(iter) *iter);
+
+L__SC void L__(init_iter)(L__(iter) *iter);
+
+#define AVL_IMPL_INIT			1
+#define AVL_IMPL_IS_EMPTY		(1 << 1)
+#define AVL_IMPL_INSERT			(1 << 2)
+#define AVL_IMPL_SEARCH			(1 << 3)
+#define AVL_IMPL_SEARCH_LEAST		(1 << 4)
+#define AVL_IMPL_SEARCH_GREATEST	(1 << 5)
+#define AVL_IMPL_REMOVE			(1 << 6)
+#define AVL_IMPL_BUILD			(1 << 7)
+#define AVL_IMPL_START_ITER		(1 << 8)
+#define AVL_IMPL_START_ITER_LEAST	(1 << 9)
+#define AVL_IMPL_START_ITER_GREATEST	(1 << 10)
+#define AVL_IMPL_GET_ITER		(1 << 11)
+#define AVL_IMPL_INCR_ITER		(1 << 12)
+#define AVL_IMPL_DECR_ITER		(1 << 13)
+#define AVL_IMPL_INIT_ITER		(1 << 14)
+#define AVL_IMPL_SUBST			(1 << 15)
+
+#define AVL_IMPL_ALL			(~0)
+
+#undef L__
+#undef L__EST_LONG_BIT
+#undef L__SIZE
+#undef L__SC
+#undef L__LONG_BIT
+#undef L__BIT_ARR_DEFN

servers/vm/fork.c

@@ -13,6 +13,7 @@
 #include <minix/ipc.h>
 #include <minix/sysutil.h>
 #include <minix/syslib.h>
+#include <minix/debug.h>
 
 #include <errno.h>
 #include <env.h>
@@ -31,6 +32,8 @@ PUBLIC int do_fork(message *msg)
 {
   int r, proc, s, childproc, fullvm;
   struct vmproc *vmp, *vmc;
+  pt_t origpt;
+  vir_bytes msgaddr;
 
   SANITYCHECK(SCL_FUNCTIONS);
 
@@ -49,6 +52,7 @@ PUBLIC int do_fork(message *msg)
 
   vmp = &vmproc[proc];		/* parent */
   vmc = &vmproc[childproc];	/* child */
+  vm_assert(vmc->vm_slot == childproc);
 
   if(vmp->vm_flags & VMF_HAS_DMA) {
 	printf("VM: %d has DMA memory and may not fork\n", msg->VMF_ENDPOINT);
@@ -58,14 +62,20 @@ PUBLIC int do_fork(message *msg)
   fullvm = vmp->vm_flags & VMF_HASPT;
 
   /* The child is basically a copy of the parent. */
+  origpt = vmc->vm_pt;
   *vmc = *vmp;
+  vmc->vm_slot = childproc;
   vmc->vm_regions = NULL;
   vmc->vm_endpoint = NONE;	/* In case someone tries to use it. */
+  vmc->vm_pt = origpt;
+  vmc->vm_flags &= ~VMF_HASPT;
 
 #if VMSTATS
   vmc->vm_bytecopies = 0;
 #endif
 
+  SANITYCHECK(SCL_DETAIL);
+
   if(fullvm) {
 	SANITYCHECK(SCL_DETAIL);
 
@@ -74,6 +84,8 @@ PUBLIC int do_fork(message *msg)
 		return ENOMEM;
 	}
 
+	vmc->vm_flags |= VMF_HASPT;
+
 	SANITYCHECK(SCL_DETAIL);
 
 	if(map_proc_copy(vmc, vmp) != OK) {
@@ -108,6 +120,7 @@ PUBLIC int do_fork(message *msg)
 	/* Create a copy of the parent's core image for the child. */
 	child_abs = (phys_bytes) child_base << CLICK_SHIFT;
 	parent_abs = (phys_bytes) vmp->vm_arch.vm_seg[D].mem_phys << CLICK_SHIFT;
+	FIXME("VM uses kernel for abscopy");
 	s = sys_abscopy(parent_abs, child_abs, prog_bytes);
 	if (s < 0) vm_panic("do_fork can't copy", s);
 
@@ -127,11 +140,18 @@ PUBLIC int do_fork(message *msg)
   /* Tell kernel about the (now successful) FORK. */
   if((r=sys_fork(vmp->vm_endpoint, childproc,
 	&vmc->vm_endpoint, vmc->vm_arch.vm_seg,
-	fullvm ? PFF_VMINHIBIT : 0)) != OK) {
+	fullvm ? PFF_VMINHIBIT : 0, &msgaddr)) != OK) {
         vm_panic("do_fork can't sys_fork", r);
   }
 
   if(fullvm) {
+	vir_bytes vir;
+	vir = arch_vir2map(vmc, msgaddr);
+	if(handle_memory(vmc, vir, sizeof(message), 1) != OK)
+		vm_panic("can't make message writable (child)", NO_NUM);
+	vir = arch_vir2map(vmp, msgaddr);
+	if(handle_memory(vmp, vir, sizeof(message), 1) != OK)
+		vm_panic("can't make message writable (parent)", NO_NUM);
 	if((r=pt_bind(&vmc->vm_pt, vmc)) != OK)
 		vm_panic("fork can't pt_bind", r);
   }

servers/vm/glo.h

@@ -25,4 +25,5 @@ EXTERN long vm_sanitychecklevel;
 
 /* vm operation mode state and values */
 EXTERN long vm_paged;
-EXTERN phys_bytes kernel_top_bytes;
+
+EXTERN int meminit_done;

servers/vm/i386/memory.h

@@ -15,7 +15,7 @@
 #define VM_PAGE_SIZE	I386_PAGE_SIZE
 
 /* Where do processes start in linear (i.e. page table) memory? */
-#define VM_PROCSTART	(I386_BIG_PAGE_SIZE*10)
+#define VM_PROCSTART	(I386_BIG_PAGE_SIZE*100)
 
 #define CLICKSPERPAGE (I386_PAGE_SIZE/CLICK_SIZE)
 

servers/vm/i386/pagetable.c

@@ -34,14 +34,14 @@
 
 #include "memory.h"
 
-int global_bit_ok = 0;
-int bigpage_ok = 0;
+/* PDE used to map in kernel, kernel physical address. */
+PRIVATE int kernel_pde = -1, pagedir_pde = -1;
+PRIVATE u32_t kern_pde_val = 0, global_bit = 0, pagedir_pde_val;
 
-/* Location in our virtual address space where we can map in 
- * any physical page we want.
-*/
-static unsigned char *varmap = NULL;	/* Our address space. */
-static u32_t varmap_loc;		/* Our page table. */
+PRIVATE int proc_pde = 0;
+
+/* 4MB page size available in hardware? */
+PRIVATE int bigpage_ok = 0;
 
 /* Our process table entry. */
 struct vmproc *vmp = &vmproc[VM_PROC_NR];
@@ -52,7 +52,7 @@ struct vmproc *vmp = &vmproc[VM_PROC_NR];
  */
 #define SPAREPAGES 5
 int missing_spares = SPAREPAGES;
-static struct {
+PRIVATE struct {
 	void *page;
 	u32_t phys;
 } sparepages[SPAREPAGES];
@@ -78,7 +78,6 @@ static struct {
 u32_t page_directories_phys, *page_directories = NULL;
 
 #if SANITYCHECKS
-#define PT_SANE(p) { pt_sanitycheck((p), __FILE__, __LINE__); SANITYCHECK(SCL_DETAIL); }
 /*===========================================================================*
  *				pt_sanitycheck		     		     *
  *===========================================================================*/
@@ -86,21 +85,33 @@ PUBLIC void pt_sanitycheck(pt_t *pt, char *file, int line)
 {
 /* Basic pt sanity check. */
 	int i;
+	int slot;
 
 	MYASSERT(pt);
 	MYASSERT(pt->pt_dir);
 	MYASSERT(pt->pt_dir_phys);
 
-	for(i = 0; i < I386_VM_DIR_ENTRIES; i++) {
+	for(slot = 0; slot < ELEMENTS(vmproc); slot++) {
+		if(pt == &vmproc[slot].vm_pt)
+			break;
+	}
+
+	if(slot >= ELEMENTS(vmproc)) {
+		vm_panic("pt_sanitycheck: passed pt not in any proc", NO_NUM);
+	}
+
+	for(i = proc_pde; i < I386_VM_DIR_ENTRIES; i++) {
 		if(pt->pt_pt[i]) {
+			if(!(pt->pt_dir[i] & I386_VM_PRESENT)) {
+				printf("slot %d: pt->pt_pt[%d] = 0x%lx, but pt_dir entry 0x%lx\n",
+					slot, i, pt->pt_pt[i], pt->pt_dir[i]);
+			}
 			MYASSERT(pt->pt_dir[i] & I386_VM_PRESENT);
 		} else {
 			MYASSERT(!(pt->pt_dir[i] & I386_VM_PRESENT));
 		}
 	}
 }
-#else
-#define PT_SANE(p)
 #endif
 
 /*===========================================================================*
@@ -262,10 +273,7 @@ PRIVATE void *vm_checkspares(void)
 	}
 	if(worst < n) worst = n;
 	total += n;
-#if 0
-	if(n > 0)
-		printf("VM: made %d spares, total %d, worst %d\n", n, total, worst);
-#endif
+
 	return NULL;
 }
 
@@ -293,7 +301,7 @@ PUBLIC void *vm_allocpages(phys_bytes *phys, int pages, int reason)
 	vm_assert(level >= 1);
 	vm_assert(level <= 2);
 
-	if(level > 1 || !(vmp->vm_flags & VMF_HASPT)) {
+	if(level > 1 || !(vmp->vm_flags & VMF_HASPT) || !meminit_done) {
 		int r;
 		void *s;
 		vm_assert(pages == 1);
@@ -347,7 +355,7 @@ PRIVATE int pt_ptalloc(pt_t *pt, int pde, u32_t flags)
 
 	/* Argument must make sense. */
 	vm_assert(pde >= 0 && pde < I386_VM_DIR_ENTRIES);
-	vm_assert(!(flags & ~(PTF_ALLFLAGS | PTF_MAPALLOC)));
+	vm_assert(!(flags & ~(PTF_ALLFLAGS)));
 
 	/* We don't expect to overwrite page directory entry, nor
 	 * storage for the page table.
@@ -385,10 +393,9 @@ PUBLIC int pt_writemap(pt_t *pt, vir_bytes v, phys_bytes physaddr,
 /* Page directory and table entries for this virtual address. */
 	int p, pages, pde;
 	int finalpde;
-	SANITYCHECK(SCL_FUNCTIONS);
 
 	vm_assert(!(bytes % I386_PAGE_SIZE));
-	vm_assert(!(flags & ~(PTF_ALLFLAGS | PTF_MAPALLOC)));
+	vm_assert(!(flags & ~(PTF_ALLFLAGS)));
 
 	pages = bytes / I386_PAGE_SIZE;
 
@@ -417,6 +424,8 @@ PUBLIC int pt_writemap(pt_t *pt, vir_bytes v, phys_bytes physaddr,
 	for(pde = I386_VM_PDE(v); pde <= finalpde; pde++) {
 		vm_assert(pde >= 0 && pde < I386_VM_DIR_ENTRIES);
 		if(pt->pt_dir[pde] & I386_VM_BIGPAGE) {
+			printf("pt_writemap: trying to write 0x%lx into 0x%lx\n",
+				physaddr, v);
                         vm_panic("pt_writemap: BIGPAGE found", NO_NUM);
 		}
 		if(!(pt->pt_dir[pde] & I386_VM_PRESENT)) {
@@ -470,7 +479,7 @@ PUBLIC int pt_writemap(pt_t *pt, vir_bytes v, phys_bytes physaddr,
 		v += I386_PAGE_SIZE;
 	PT_SANE(pt);
 	}
-	SANITYCHECK(SCL_FUNCTIONS);
+
 	PT_SANE(pt);
 
 	return OK;
@@ -488,7 +497,14 @@ PUBLIC int pt_new(pt_t *pt)
  */
 	int i;
 
-        if(!(pt->pt_dir = vm_allocpages(&pt->pt_dir_phys, 1, VMP_PAGEDIR))) {
+	/* Don't ever re-allocate/re-move a certain process slot's
+	 * page directory once it's been created. This is a fraction
+	 * faster, but also avoids having to invalidate the page
+	 * mappings from in-kernel page tables pointing to
+	 * the page directories (the page_directories data).
+	 */
+        if(!pt->pt_dir &&
+          !(pt->pt_dir = vm_allocpages(&pt->pt_dir_phys, 1, VMP_PAGEDIR))) {
 		return ENOMEM;
 	}
 
@@ -500,10 +516,14 @@ PUBLIC int pt_new(pt_t *pt)
 	/* Where to start looking for free virtual address space? */
 	pt->pt_virtop = 0;
 
+	PT_SANE(pt);
+
         /* Map in kernel. */
         if(pt_mapkernel(pt) != OK)
                 vm_panic("pt_new: pt_mapkernel failed", NO_NUM);
 
+	PT_SANE(pt);
+
 	return OK;
 }
 
@@ -520,13 +540,14 @@ PUBLIC void pt_init(void)
  */
         pt_t *newpt;
         int s, r;
-        vir_bytes v;
+        vir_bytes v, kpagedir;
         phys_bytes lo, hi; 
         vir_bytes extra_clicks;
         u32_t moveup = 0;
-
-	global_bit_ok = _cpufeature(_CPUF_I386_PGE);
-	bigpage_ok = _cpufeature(_CPUF_I386_PSE);
+	int global_bit_ok = 0;
+	int free_pde;
+	int p;
+	vir_bytes kernlimit;
 
         /* Shorthand. */
         newpt = &vmp->vm_pt;
@@ -542,11 +563,36 @@ PUBLIC void pt_init(void)
 
 	missing_spares = 0;
 
-        /* Make new page table for ourselves, partly copied
-         * from the current one.
+	/* global bit and 4MB pages available? */
+	global_bit_ok = _cpufeature(_CPUF_I386_PGE);
+	bigpage_ok = _cpufeature(_CPUF_I386_PSE);
+
+	/* Set bit for PTE's and PDE's if available. */
+	if(global_bit_ok)
+		global_bit = I386_VM_GLOBAL;
+
+	/* Figure out kernel pde slot. */
+	{
+		int pde1, pde2;
+		pde1 = I386_VM_PDE(KERNEL_TEXT);
+		pde2 = I386_VM_PDE(KERNEL_DATA+KERNEL_DATA_LEN);
+		if(pde1 != pde2)
+                	vm_panic("pt_init: kernel too big", NO_NUM); 
+
+		/* Map in kernel with this single pde value if 4MB pages
+		 * supported.
 		 */
-        if(pt_new(newpt) != OK)
-                vm_panic("pt_init: pt_new failed", NO_NUM); 
+		kern_pde_val = (KERNEL_TEXT & I386_VM_ADDR_MASK_4MB) |
+				I386_VM_BIGPAGE|
+				I386_VM_USER|
+				I386_VM_PRESENT|I386_VM_WRITE|global_bit;
+		kernel_pde = pde1;
+		vm_assert(kernel_pde >= 0);
+		free_pde = kernel_pde+1;
+	}
+
+	/* First unused pde. */
+	proc_pde = free_pde;
            
         /* Initial (current) range of our virtual address space. */
         lo = CLICK2ABS(vmp->vm_arch.vm_seg[T].mem_phys);
@@ -563,20 +609,26 @@ PUBLIC void pt_init(void)
                 vm_assert(!(moveup % I386_PAGE_SIZE));
         }
         
+        /* Make new page table for ourselves, partly copied
+         * from the current one.
+         */     
+        if(pt_new(newpt) != OK)
+                vm_panic("pt_init: pt_new failed", NO_NUM); 
+
+	/* Old position mapped in? */
+	pt_check(vmp);
+                
         /* Set up mappings for VM process. */
         for(v = lo; v < hi; v += I386_PAGE_SIZE)  {
                 phys_bytes addr;
                 u32_t flags; 
         
-                /* We have to write the old and new position in the PT,
+                /* We have to write the new position in the PT,
                  * so we can move our segments.
                  */ 
                 if(pt_writemap(newpt, v+moveup, v, I386_PAGE_SIZE,
                         I386_VM_PRESENT|I386_VM_WRITE|I386_VM_USER, 0) != OK)
                         vm_panic("pt_init: pt_writemap failed", NO_NUM);
-                if(pt_writemap(newpt, v, v, I386_PAGE_SIZE,
-                        I386_VM_PRESENT|I386_VM_WRITE|I386_VM_USER, 0) != OK)
-                        vm_panic("pt_init: pt_writemap failed", NO_NUM);
         }
        
         /* Move segments up too. */
@@ -584,21 +636,14 @@ PUBLIC void pt_init(void)
         vmp->vm_arch.vm_seg[D].mem_phys += ABS2CLICK(moveup);
         vmp->vm_arch.vm_seg[S].mem_phys += ABS2CLICK(moveup);
        
-#if 0
-        /* Map in kernel. */
-        if(pt_mapkernel(newpt) != OK)
-                vm_panic("pt_init: pt_mapkernel failed", NO_NUM);
-
 	/* Allocate us a page table in which to remember page directory
 	 * pointers.
 	 */
 	if(!(page_directories = vm_allocpages(&page_directories_phys,
 		1, VMP_PAGETABLE)))
                 vm_panic("no virt addr for vm mappings", NO_NUM);
-#endif
 
-        /* Give our process the new, copied, private page table. */
-        pt_bind(newpt, vmp);
+	memset(page_directories, 0, I386_PAGE_SIZE);
        
         /* Increase our hardware data segment to create virtual address
          * space above our stack. We want to increase it to VM_DATATOP,
@@ -614,19 +659,6 @@ PUBLIC void pt_init(void)
                 (vmp->vm_arch.vm_seg[S].mem_vir +
                 vmp->vm_arch.vm_seg[S].mem_len) << CLICK_SHIFT;
        
-        if((s=sys_newmap(VM_PROC_NR, vmp->vm_arch.vm_seg)) != OK)
-                vm_panic("VM: pt_init: sys_newmap failed", s);
-       
-        /* Back to reality - this is where the stack actually is. */
-        vmp->vm_arch.vm_seg[S].mem_len -= extra_clicks;
-       
-        /* Wipe old mappings from VM. */
-        for(v = lo; v < hi; v += I386_PAGE_SIZE)  {
-                if(pt_writemap(newpt, v, MAP_NONE, I386_PAGE_SIZE,
-                        0, WMF_OVERWRITE) != OK)
-                        vm_panic("pt_init: pt_writemap failed", NO_NUM);
-        }
-       
         /* Where our free virtual address space starts.
          * This is only a hint to the VM system.
          */
@@ -635,16 +667,48 @@ PUBLIC void pt_init(void)
         /* Let other functions know VM now has a private page table. */
         vmp->vm_flags |= VMF_HASPT;
 
-        /* Reserve a page in our virtual address space that we
-         * can use to map in arbitrary physical pages.
+	/* Find a PDE below processes available for mapping in the
+	 * page directories (readonly).
 	 */
-        varmap_loc = findhole(newpt, I386_PAGE_SIZE,
-                arch_vir2map(vmp, vmp->vm_stacktop),
-                vmp->vm_arch.vm_data_top);
-        if(varmap_loc == NO_MEM) {
-                vm_panic("no virt addr for vm mappings", NO_NUM);
+	pagedir_pde = free_pde++;
+	pagedir_pde_val = (page_directories_phys & I386_VM_ADDR_MASK) |
+			I386_VM_PRESENT | I386_VM_USER | I386_VM_WRITE;
+
+	/* Tell kernel about free pde's. */
+	while(free_pde*I386_BIG_PAGE_SIZE < VM_PROCSTART) {
+		if((r=sys_vmctl(SELF, VMCTL_I386_FREEPDE, free_pde++)) != OK) {
+			vm_panic("VMCTL_I386_FREEPDE failed", r);
 		}
-        varmap = (unsigned char *) arch_map2vir(vmp, varmap_loc);
+	}
+
+	/* first pde in use by process. */
+	proc_pde = free_pde;
+
+	kernlimit = free_pde*I386_BIG_PAGE_SIZE;
+
+	/* Increase kernel segment to address this memory. */
+	if((r=sys_vmctl(SELF, VMCTL_I386_KERNELLIMIT, kernlimit)) != OK) {
+                vm_panic("VMCTL_I386_KERNELLIMIT failed", r);
+	}
+
+	kpagedir = arch_map2vir(&vmproc[VMP_SYSTEM],
+		pagedir_pde*I386_BIG_PAGE_SIZE);
+
+	/* Tell kernel how to get at the page directories. */
+	if((r=sys_vmctl(SELF, VMCTL_I386_PAGEDIRS, kpagedir)) != OK) {
+                vm_panic("VMCTL_I386_KERNELLIMIT failed", r);
+	}
+       
+        /* Give our process the new, copied, private page table. */
+	pt_mapkernel(newpt);	/* didn't know about vm_dir pages earlier */
+        pt_bind(newpt, vmp);
+       
+	/* Now actually enable paging. */
+	if(sys_vmctl_enable_paging(vmp->vm_arch.vm_seg) != OK)
+        	vm_panic("pt_init: enable paging failed", NO_NUM);
+
+        /* Back to reality - this is where the stack actually is. */
+        vmp->vm_arch.vm_seg[S].mem_len -= extra_clicks;
        
         /* All OK. */
         return;
@@ -656,7 +720,8 @@ PUBLIC void pt_init(void)
  *===========================================================================*/
 PUBLIC int pt_bind(pt_t *pt, struct vmproc *who)
 {
-	int slot;
+	int slot, ispt;
+	u32_t phys;
 
 	/* Basic sanity checks. */
 	vm_assert(who);
@@ -664,16 +729,20 @@ PUBLIC int pt_bind(pt_t *pt, struct vmproc *who)
 	if(pt) PT_SANE(pt);
 	vm_assert(pt);
 
-#if 0
 	slot = who->vm_slot;
 	vm_assert(slot >= 0);
 	vm_assert(slot < ELEMENTS(vmproc));
-	vm_assert(!(pt->pt_dir_phys & ~I386_VM_ADDR_MASK));
+	vm_assert(slot < I386_VM_PT_ENTRIES);
 
-	page_directories[slot] = (pt->pt_dir_phys & I386_VM_ADDR_MASK) | 
-		(I386_VM_PRESENT|I386_VM_WRITE);
+	phys = pt->pt_dir_phys & I386_VM_ADDR_MASK;
+	vm_assert(pt->pt_dir_phys == phys);
+
+	/* Update "page directory pagetable." */
+	page_directories[slot] = phys | I386_VM_PRESENT|I386_VM_WRITE;
+
+#if 0
+	printf("VM: slot %d has pde val 0x%lx\n", slot, page_directories[slot]);
 #endif
-
 	/* Tell kernel about new page table root. */
 	return sys_vmctl(who->vm_endpoint, VMCTL_I386_SETCR3,
 		pt ? pt->pt_dir_phys : 0);
@@ -689,22 +758,14 @@ PUBLIC void pt_free(pt_t *pt)
 
 	PT_SANE(pt);
 
-	for(i = 0; i < I386_VM_DIR_ENTRIES; i++) {
-		int p;
-		if(pt->pt_pt[i]) {
-		   for(p = 0; p < I386_VM_PT_ENTRIES; p++) {
-			if((pt->pt_pt[i][p] & (PTF_MAPALLOC | I386_VM_PRESENT)) 
-			 == (PTF_MAPALLOC | I386_VM_PRESENT)) {
-					u32_t pa = I386_VM_PFA(pt->pt_pt[i][p]);
-					FREE_MEM(ABS2CLICK(pa), CLICKSPERPAGE);
-				}
-		  }
+	for(i = 0; i < I386_VM_DIR_ENTRIES; i++)
+		if(pt->pt_pt[i])
 			vm_freepages((vir_bytes) pt->pt_pt[i],
 				I386_VM_PFA(pt->pt_dir[i]), 1, VMP_PAGETABLE);
-		}
-	}
 
+#if 0
 	vm_freepages((vir_bytes) pt->pt_dir, pt->pt_dir_phys, 1, VMP_PAGEDIR);
+#endif
 
 	return;
 }
@@ -715,79 +776,53 @@ PUBLIC void pt_free(pt_t *pt)
 PUBLIC int pt_mapkernel(pt_t *pt)
 {
 	int r;
-	static int pde = -1, do_bigpage = 0;
-	u32_t global = 0;
-	static u32_t kern_phys;
 	static int printed = 0;
 
-	if(global_bit_ok) global = I386_VM_GLOBAL;
-
         /* Any i386 page table needs to map in the kernel address space. */
         vm_assert(vmproc[VMP_SYSTEM].vm_flags & VMF_INUSE);
 
-	if(pde == -1 && bigpage_ok) {
-		int pde1, pde2;
-		pde1 = I386_VM_PDE(KERNEL_TEXT);
-		pde2 = I386_VM_PDE(KERNEL_DATA+KERNEL_DATA_LEN);
-		if(pde1 != pde2) {
-			printf("VM: pt_mapkernel: kernel too big?");
-			bigpage_ok = 0;
+	if(bigpage_ok) {
+		if(kernel_pde >= 0) {
+			pt->pt_dir[kernel_pde] = kern_pde_val;
+		} else
+			vm_panic("VM: pt_mapkernel: no kernel pde", NO_NUM);
 	} else {
-			kern_phys = KERNEL_TEXT & I386_VM_ADDR_MASK_4MB;
-			pde = pde1;
-			do_bigpage = 1;
-			vm_assert(pde >= 0);
-		}
-	}
+		vm_panic("VM: pt_mapkernel: no bigpage", NO_NUM);
 
-	if(do_bigpage) {
-		pt->pt_dir[pde] = kern_phys |
-			I386_VM_BIGPAGE|I386_VM_PRESENT|I386_VM_WRITE|global;
-	} else {
         	/* Map in text. flags: don't write, supervisor only */
         	if((r=pt_writemap(pt, KERNEL_TEXT, KERNEL_TEXT, KERNEL_TEXT_LEN,
-			I386_VM_PRESENT|global, 0)) != OK)
+			I386_VM_PRESENT|global_bit, 0)) != OK)
 			return r;
  
         	/* Map in data. flags: read-write, supervisor only */
         	if((r=pt_writemap(pt, KERNEL_DATA, KERNEL_DATA, KERNEL_DATA_LEN,
-			I386_VM_PRESENT|I386_VM_WRITE|global, 0)) != OK)
+			I386_VM_PRESENT|I386_VM_WRITE, 0)) != OK)
 			return r;
 	}
 
+	if(pagedir_pde >= 0) {
+		/* Kernel also wants to know about all page directories. */
+		pt->pt_dir[pagedir_pde] = pagedir_pde_val;
+	}
+
 	return OK;
 }
 
 /*===========================================================================*
- *				pt_freerange		     		     *
+ *                              pt_check                                *
  *===========================================================================*/
-PUBLIC void pt_freerange(pt_t *pt, vir_bytes low, vir_bytes high)
+PUBLIC void pt_check(struct vmproc *vmp)
 {
-/* Free memory allocated by pagetable functions in this range. */
-	int pde;
-	u32_t v;
-
-	PT_SANE(pt);
-
-	for(v = low; v < high; v += I386_PAGE_SIZE) {
-		int pte;
-		pde = I386_VM_PDE(v);
-		pte = I386_VM_PTE(v);
-		if(!(pt->pt_dir[pde] & I386_VM_PRESENT))
-			continue;
-		if((pt->pt_pt[pde][pte] & (PTF_MAPALLOC | I386_VM_PRESENT)) 
-		 == (PTF_MAPALLOC | I386_VM_PRESENT)) {
-			u32_t pa = I386_VM_PFA(pt->pt_pt[pde][pte]);
-			FREE_MEM(ABS2CLICK(pa), CLICKSPERPAGE);
-			pt->pt_pt[pde][pte] = 0;
+	phys_bytes hi;
+        hi = CLICK2ABS(vmp->vm_arch.vm_seg[S].mem_phys +
+                vmp->vm_arch.vm_seg[S].mem_len);
+	if(hi >= (kernel_pde+1) * I386_BIG_PAGE_SIZE) {
+		printf("VM: %d doesn't fit in kernel range\n",
+			vmp->vm_endpoint, hi);
+		vm_panic("boot time processes too big", NO_NUM);
 	}
 }
 
-	PT_SANE(pt);
-
-	return;
-}
-
 /*===========================================================================*
  *				pt_cycle		     		     *
  *===========================================================================*/
@@ -796,82 +831,3 @@ PUBLIC void pt_cycle(void)
 	vm_checkspares();
 }
 
-/* In sanity check mode, pages are mapped and unmapped explicitly, so
- * unexpected double mappings (overwriting a page table entry) are caught.
- * If not sanity checking, simply keep the page mapped in and overwrite
- * the mapping entry; we need WMF_OVERWRITE for that in PHYS_MAP though.
- */
-#if SANITYCHECKS
-#define MAPFLAGS	0
-#else
-#define MAPFLAGS	WMF_OVERWRITE
-#endif
-
-static u32_t ismapped = MAP_NONE;
-
-#define PHYS_MAP(a, o)							\
-{	int r;								\
-	u32_t wantmapped;						\
-	vm_assert(varmap);						\
-	(o) = (a) % I386_PAGE_SIZE;					\
-	wantmapped = (a) - (o);						\
-	if(wantmapped != ismapped || ismapped == MAP_NONE) {		\
-		r = pt_writemap(&vmp->vm_pt, (vir_bytes) varmap_loc, 	\
-			wantmapped, I386_PAGE_SIZE, 			\
-			I386_VM_PRESENT | I386_VM_USER | I386_VM_WRITE, \
-				MAPFLAGS); 				\
-		if(r != OK)						\
-			vm_panic("PHYS_MAP: pt_writemap", NO_NUM);	\
-		ismapped = wantmapped;					\
-		/* pt_bind() flushes TLB. */				\
-		pt_bind(&vmp->vm_pt, vmp); 				\
-	}								\
-}
-
-#define PHYSMAGIC 0x7b9a0590
-
-#if SANITYCHECKS
-#define PHYS_UNMAP if(OK != pt_writemap(&vmp->vm_pt, varmap_loc, MAP_NONE,\
-	I386_PAGE_SIZE, 0, WMF_OVERWRITE)) {				\
-		vm_panic("PHYS_UNMAP: pt_writemap failed", NO_NUM); }
-	ismapped = MAP_NONE;
-#endif
-
-#define PHYS_VAL(o) (* (phys_bytes *) (varmap + (o)))
-
-
-/*===========================================================================*
- *                              phys_writeaddr                               *
- *===========================================================================*/
-PUBLIC void phys_writeaddr(phys_bytes addr, phys_bytes v1, phys_bytes v2)
-{
-	phys_bytes offset;
-
-	SANITYCHECK(SCL_DETAIL);
-	PHYS_MAP(addr, offset);
-	PHYS_VAL(offset) = v1;
-	PHYS_VAL(offset + sizeof(phys_bytes)) = v2;
-#if SANITYCHECKS
-	PHYS_VAL(offset + 2*sizeof(phys_bytes)) = PHYSMAGIC;
-	PHYS_UNMAP;
-#endif
-	SANITYCHECK(SCL_DETAIL);
-}
-
-/*===========================================================================*
- *                              phys_readaddr                                *
- *===========================================================================*/
-PUBLIC void phys_readaddr(phys_bytes addr, phys_bytes *v1, phys_bytes *v2)
-{
-	phys_bytes offset;
-
-	SANITYCHECK(SCL_DETAIL);
-	PHYS_MAP(addr, offset);
-	*v1 = PHYS_VAL(offset);
-	*v2 = PHYS_VAL(offset + sizeof(phys_bytes));
-#if SANITYCHECKS
-	vm_assert(PHYS_VAL(offset + 2*sizeof(phys_bytes)) == PHYSMAGIC);
-	PHYS_UNMAP;
-#endif
-	SANITYCHECK(SCL_DETAIL);
-}

servers/vm/i386/pagetable.h

@@ -5,6 +5,8 @@
 #include <stdint.h>
 #include <sys/vm_i386.h>
 
+#include "../vm.h"
+
 /* An i386 pagetable. */
 typedef struct {
 	/* Directory entries in VM addr space - root of page table.  */
@@ -34,5 +36,12 @@ typedef struct {
  */
 #define PTF_ALLFLAGS	(PTF_WRITE|PTF_PRESENT|PTF_USER|PTF_GLOBAL)
 
+#if SANITYCHECKS
+#define PT_SANE(p) { pt_sanitycheck((p), __FILE__, __LINE__); }
+#else
+#define PT_SANE(p)
 #endif
 
+#endif
+
+

servers/vm/i386/vm.c

@@ -25,83 +25,18 @@
 
 #include "memory.h"
 
-#define PAGE_SIZE	4096
-#define PAGE_DIR_SIZE	(1024*PAGE_SIZE)	
-#define PAGE_TABLE_COVER (1024*PAGE_SIZE)
-/*=========================================================================*
- *				arch_init_vm				   *
- *=========================================================================*/
-PUBLIC void arch_init_vm(mem_chunks)
-struct memory mem_chunks[NR_MEMS];
-{
-	phys_bytes high, bytes;
-	phys_clicks clicks, base_click;
-	unsigned pages;
-	int i, r;
-
-	/* Compute the highest memory location */
-	high= 0;
-	for (i= 0; i<NR_MEMS; i++)
-	{
-		if (mem_chunks[i].size == 0)
-			continue;
-		if (mem_chunks[i].base + mem_chunks[i].size > high)
-			high= mem_chunks[i].base + mem_chunks[i].size;
-	}
-
-	high <<= CLICK_SHIFT;
-#if VERBOSE_VM
-	printf("do_x86_vm: found high 0x%x\n", high);
-#endif
-	
-	/* Rounding up */
-	high= (high-1+PAGE_DIR_SIZE) & ~(PAGE_DIR_SIZE-1);
-
-	/* The number of pages we need is one for the page directory, enough
-	 * page tables to cover the memory, and one page for alignement.
-	 */
-	pages= 1 + (high + PAGE_TABLE_COVER-1)/PAGE_TABLE_COVER + 1;
-	bytes= pages*PAGE_SIZE;
-	clicks= (bytes + CLICK_SIZE-1) >> CLICK_SHIFT;
-
-#if VERBOSE_VM
-	printf("do_x86_vm: need %d pages\n", pages);
-	printf("do_x86_vm: need %d bytes\n", bytes);
-	printf("do_x86_vm: need %d clicks\n", clicks);
-#endif
-
-	for (i= 0; i<NR_MEMS; i++)
-	{
-		if (mem_chunks[i].size <= clicks)
-			continue;
-		break;
-	}
-	if (i >= NR_MEMS)
-		panic("VM", "not enough memory for VM page tables?", NO_NUM);
-	base_click= mem_chunks[i].base;
-	mem_chunks[i].base += clicks;
-	mem_chunks[i].size -= clicks;
-
-#if VERBOSE_VM
-	printf("do_x86_vm: using 0x%x clicks @ 0x%x\n", clicks, base_click);
-#endif
-	r= sys_vm_setbuf(base_click << CLICK_SHIFT, clicks << CLICK_SHIFT,
-		high);
-	if (r != 0)
-		printf("do_x86_vm: sys_vm_setbuf failed: %d\n", r);
-
-}
-
 /*===========================================================================*
  *				arch_map2vir				     *
  *===========================================================================*/
 PUBLIC vir_bytes arch_map2vir(struct vmproc *vmp, vir_bytes addr)
 {
-	vir_bytes bottom = CLICK2ABS(vmp->vm_arch.vm_seg[D].mem_phys);
+	vir_bytes textstart = CLICK2ABS(vmp->vm_arch.vm_seg[T].mem_phys);
+	vir_bytes datastart = CLICK2ABS(vmp->vm_arch.vm_seg[D].mem_phys);
 
-	vm_assert(bottom <= addr);
+	/* Could be a text address. */
+	vm_assert(datastart <= addr || textstart <= addr);
 
-	return addr - bottom;
+	return addr - datastart;
 }
 
 /*===========================================================================*

servers/vm/main.c

@@ -77,6 +77,7 @@ PUBLIC int main(void)
 
 #if SANITYCHECKS
   nocheck = 0;
+  FIXME("VM SANITYCHECKS are on");
   memcpy(data1, CHECKADDR, sizeof(data1));    
 #endif
 	SANITYCHECK(SCL_TOP);
@@ -114,21 +115,18 @@ PUBLIC int main(void)
 		switch(msg.m_source) {
 			case SYSTEM:
 				/* Kernel wants to have memory ranges
-				 * verified.
+				 * verified, and/or pagefaults handled.
 				 */
 				do_memory();
 				break;
+			case HARDWARE:
+				do_pagefaults();
+				break;
 			case PM_PROC_NR:
 				/* PM sends a notify() on shutdown, which
 				 * is OK and we ignore.
 				 */
 				break;
-			case HARDWARE:
-				/* This indicates a page fault has happened,
-				 * which we have to handle.
-				 */
-				do_pagefaults();
-				break;
 			default:
 				/* No-one else should send us notifies. */
 				printf("VM: ignoring notify() from %d\n",
@@ -177,6 +175,7 @@ PUBLIC int main(void)
 PRIVATE void vm_init(void)
 {
 	int s, i;
+	int click, clicksforgotten = 0;
 	struct memory mem_chunks[NR_MEMS];
 	struct boot_image image[NR_BOOT_PROCS];
 	struct boot_image *ip;
@@ -241,37 +240,12 @@ PRIVATE void vm_init(void)
 			vmp->vm_flags |= VMF_SEPARATE;
 	}
 
-
-	/* Let architecture-dependent VM initialization use some memory. */
-	arch_init_vm(mem_chunks);
-
 	/* Architecture-dependent initialization. */
 	pt_init();
 
 	/* Initialize tables to all physical memory. */
 	mem_init(mem_chunks);
-
-	/* Bits of code need to know where a process can
-	 * start in a pagetable.
-	 */
-        kernel_top_bytes = find_kernel_top();
-
-	/* Can first kernel pages of code and data be (left) mapped out?
-	 * If so, change the SYSTEM process' memory map to reflect this
-	 * (future mappings of SYSTEM into other processes will not include
-	 * first pages), and free the first pages.
-	 */
-	if(vm_paged && sys_vmctl(SELF, VMCTL_NOPAGEZERO, 0) == OK) {
-	  	struct vmproc *vmp;
-		vmp = &vmproc[VMP_SYSTEM];
-		if(vmp->vm_arch.vm_seg[T].mem_len > 0) {
-#define DIFF CLICKSPERPAGE
-			vmp->vm_arch.vm_seg[T].mem_phys += DIFF;
-			vmp->vm_arch.vm_seg[T].mem_len -= DIFF;
-		}
-		vmp->vm_arch.vm_seg[D].mem_phys += DIFF;
-		vmp->vm_arch.vm_seg[D].mem_len -= DIFF;
-	}
+	meminit_done = 1;
 
 	/* Give these processes their own page table. */
 	for (ip = &image[0]; ip < &image[NR_BOOT_PROCS]; ip++) {
@@ -283,14 +257,22 @@ PRIVATE void vm_init(void)
 
 		GETVMP(vmp, ip->proc_nr);
 
+		if(!(ip->flags & PROC_FULLVM)) {
+			/* See if this process fits in kernel
+			 * mapping. VM has its own pagetable,
+			 * don't check it.
+			 */
+			if(!(vmp->vm_flags & VMF_HASPT)) {
+				pt_check(vmp);
+			}
+			continue;
+		}
+
 		old_stack = 
 			vmp->vm_arch.vm_seg[S].mem_vir +
 			vmp->vm_arch.vm_seg[S].mem_len - 
 			vmp->vm_arch.vm_seg[D].mem_len;
 
-		if(!(ip->flags & PROC_FULLVM))
-			continue;
-
         	if(pt_new(&vmp->vm_pt) != OK)
 			vm_panic("vm_init: no new pagetable", NO_NUM);
 #define BASICSTACK VM_PAGE_SIZE
@@ -370,11 +352,3 @@ PRIVATE void vm_init(void)
 		vm_panic("kernel loaded too high", NO_NUM);
 }
 
-#if 0
-void kputc(int c)
-{
-	if(c == '\n')
-		ser_putc('\r');
-	ser_putc(c);
-}
-#endif

servers/vm/mmap.c

@@ -47,10 +47,6 @@ PUBLIC int do_mmap(message *m)
 
 	vmp = &vmproc[n];
 
-	if(m->VMM_FLAGS & MAP_LOWER16M)
-		printf("VM: warning for %d: MAP_LOWER16M not implemented\n",
-			m->m_source);
-
 	if(!(vmp->vm_flags & VMF_HASPT))
 		return ENXIO;
 
@@ -66,14 +62,15 @@ PUBLIC int do_mmap(message *m)
 
 		if(m->VMM_FLAGS & MAP_CONTIG) mfflags |= MF_CONTIG;
 		if(m->VMM_FLAGS & MAP_PREALLOC) mfflags |= MF_PREALLOC;
+		if(m->VMM_FLAGS & MAP_LOWER16M) vrflags |= VR_LOWER16MB;
 		if(m->VMM_FLAGS & MAP_ALIGN64K) vrflags |= VR_PHYS64K;
 
 		if(len % VM_PAGE_SIZE)
 			len += VM_PAGE_SIZE - (len % VM_PAGE_SIZE);
 
 		if(!(vr = map_page_region(vmp,
-			arch_vir2map(vmp, vmp->vm_stacktop), VM_DATATOP, len, MAP_NONE,
-			vrflags, mfflags))) {
+			arch_vir2map(vmp, vmp->vm_stacktop),
+			VM_DATATOP, len, MAP_NONE, vrflags, mfflags))) {
 			return ENOMEM;
 		}
 	} else {
@@ -84,6 +81,7 @@ PUBLIC int do_mmap(message *m)
 	vm_assert(vr);
 	m->VMM_RETADDR = arch_map2vir(vmp, vr->vaddr);
 
+
 	return OK;
 }
 

servers/vm/pagefaults.c

@@ -61,12 +61,21 @@ PUBLIC void do_pagefaults(void)
 		vir_bytes offset;
 		int p, wr = PFERR_WRITE(err);
 
+#if 0
+		printf("VM: pagefault: ep %d 0x%lx %s\n", 
+			ep, arch_map2vir(vmp, addr), pf_errstr(err));
+#endif
+
 		if(vm_isokendpt(ep, &p) != OK)
 			vm_panic("do_pagefaults: endpoint wrong", ep);
 
 		vmp = &vmproc[p];
 		vm_assert(vmp->vm_flags & VMF_INUSE);
 
+#if 0
+		map_printmap(vmp);
+#endif
+
 		/* See if address is valid at all. */
 		if(!(region = map_lookup(vmp, addr))) {
 			vm_assert(PFERR_NOPAGE(err));
@@ -104,11 +113,19 @@ PUBLIC void do_pagefaults(void)
 				vm_panic("sys_kill failed", s);
 			continue;
 		}
+#if 0
+		printf("VM: map_pf done; ep %d 0x%lx %s\n", 
+			ep, arch_map2vir(vmp, addr), pf_errstr(err));
 
 
+		printf("VM: handling pagefault OK: %d addr 0x%lx %s\n", 
+			ep, arch_map2vir(vmp, addr), pf_errstr(err));
+#endif
+
 		/* Pagefault is handled, so now reactivate the process. */
 		if((s=sys_vmctl(ep, VMCTL_CLEAR_PAGEFAULT, r)) != OK)
 			vm_panic("do_pagefaults: sys_vmctl failed", ep);
+
 	}
 
 	return;
@@ -120,21 +137,43 @@ PUBLIC void do_pagefaults(void)
 PUBLIC void do_memory(void)
 {
 	int r, s;
-	endpoint_t who;
+	endpoint_t who, requestor;
 	vir_bytes mem;
 	vir_bytes len;
 	int wrflag;
 
-	while((r=sys_vmctl_get_memreq(&who, &mem, &len, &wrflag)) == OK) {
+	while((r=sys_vmctl_get_memreq(&who, &mem, &len, &wrflag, &requestor))
+	  == OK) {
 		int p, r = OK;
-		struct vir_region *region;
 		struct vmproc *vmp;
-		vir_bytes o;
 
 		if(vm_isokendpt(who, &p) != OK)
 			vm_panic("do_memory: endpoint wrong", who);
 		vmp = &vmproc[p];
 
+		r = handle_memory(vmp, mem, len, wrflag);
+
+		if(sys_vmctl(requestor, VMCTL_MEMREQ_REPLY, r) != OK)
+			vm_panic("do_memory: sys_vmctl failed", r);
+
+#if 0
+		printf("VM: handling memory request %d done OK\n",
+			who);
+#endif
+	}
+}
+
+int handle_memory(struct vmproc *vmp, vir_bytes mem, vir_bytes len, int wrflag)
+{
+	struct vir_region *region;
+	vir_bytes o;
+	int r;
+
+#if 0
+	printf("VM: handling memory request: %d, 0x%lx-0x%lx, wr %d\n",
+		vmp->vm_endpoint, mem, mem+len, wrflag);
+#endif
+
 	/* Page-align memory and length. */
 	o = mem % VM_PAGE_SIZE;
 	mem -= o;
@@ -143,6 +182,7 @@ PUBLIC void do_memory(void)
 	if(o > 0) len += VM_PAGE_SIZE - o;
 
 	if(!(region = map_lookup(vmp, mem))) {
+		map_printmap(vmp);
 		printf("VM: do_memory: memory doesn't exist\n");
 		r = EFAULT;
 	} else if(mem + len > region->vaddr + region->length) {
@@ -166,9 +206,4 @@ PUBLIC void do_memory(void)
 			arch_map2vir(vmp, mem), arch_map2vir(vmp, mem+len),
 			vmp->vm_endpoint);
 	}
-
-		if(sys_vmctl(who, VMCTL_MEMREQ_REPLY, r) != OK)
-			vm_panic("do_memory: sys_vmctl failed", r);
 }
-}
-

servers/vm/pagerange.h

@@ -0,0 +1,24 @@
+
+
+#include <minix/callnr.h>
+#include <minix/com.h>
+#include <minix/config.h>
+#include <minix/const.h>
+#include <minix/ds.h>
+#include <minix/endpoint.h>
+#include <minix/keymap.h>
+#include <minix/minlib.h>
+#include <minix/type.h>
+#include <minix/ipc.h>
+#include <minix/sysutil.h>
+#include <minix/syslib.h>
+#include <minix/const.h>
+
+typedef struct pagerange {
+	phys_bytes	addr;	/* in pages */
+	phys_bytes	size;	/* in pages */
+
+	/* AVL fields */
+	struct pagerange *less, *greater;	/* children */
+	int		factor;	/* AVL balance factor */
+} pagerange_t;

servers/vm/proto.h

@@ -81,9 +81,12 @@ _PROTOTYPE(int do_unmap_phys, (message *msg)                            );
 _PROTOTYPE( void do_pagefaults, (void)				);
 _PROTOTYPE( void do_memory, (void)				);
 _PROTOTYPE( char *pf_errstr, (u32_t err));
+_PROTOTYPE( int handle_memory, (struct vmproc *vmp, vir_bytes mem,
+	vir_bytes len, int wrflag));
 
 /* $(ARCH)/pagetable.c */
 _PROTOTYPE( void pt_init, (void)					);
+_PROTOTYPE( void pt_check, (struct vmproc *vmp)				);
 _PROTOTYPE( int pt_new, (pt_t *pt)					);
 _PROTOTYPE( void pt_free, (pt_t *pt)					);
 _PROTOTYPE( void pt_freerange, (pt_t *pt, vir_bytes lo, vir_bytes hi)	);
@@ -93,8 +96,6 @@ _PROTOTYPE( int pt_bind, (pt_t *pt, struct vmproc *who)			);
 _PROTOTYPE( void *vm_allocpages, (phys_bytes *p, int pages, int cat));
 _PROTOTYPE( void pt_cycle, (void));
 _PROTOTYPE( int pt_mapkernel, (pt_t *pt));
-_PROTOTYPE( void phys_readaddr, (phys_bytes addr, phys_bytes *v1, phys_bytes *v2));
-_PROTOTYPE( void phys_writeaddr, (phys_bytes addr, phys_bytes v1, phys_bytes v2));
 #if SANITYCHECKS
 _PROTOTYPE( void pt_sanitycheck, (pt_t *pt, char *file, int line)	);
 #endif
@@ -135,6 +136,7 @@ _PROTOTYPE(int map_pf,(struct vmproc *vmp,
 	struct vir_region *region, vir_bytes offset, int write));
 _PROTOTYPE(int map_handle_memory,(struct vmproc *vmp,
 	struct vir_region *region, vir_bytes offset, vir_bytes len, int write));
+_PROTOTYPE(void map_printmap, (struct vmproc *vmp));
 
 _PROTOTYPE(struct vir_region * map_region_lookup_tag, (struct vmproc *vmp, u32_t tag));
 _PROTOTYPE(void map_region_set_tag, (struct vir_region *vr, u32_t tag));

servers/vm/region.c

@@ -32,8 +32,6 @@ FORWARD _PROTOTYPE(int map_new_physblock, (struct vmproc *vmp,
 FORWARD _PROTOTYPE(int map_copy_ph_block, (struct vmproc *vmp, struct vir_region *region, struct phys_region *ph));
 FORWARD _PROTOTYPE(struct vir_region *map_copy_region, (struct vir_region *));
 
-FORWARD _PROTOTYPE(void map_printmap, (struct vmproc *vmp));
-
 PRIVATE char *map_name(struct vir_region *vr)
 {
 	int type = vr->flags & (VR_ANON|VR_DIRECT);
@@ -52,7 +50,7 @@ PRIVATE char *map_name(struct vir_region *vr)
 /*===========================================================================*
  *				map_printmap				     *
  *===========================================================================*/
-PRIVATE void map_printmap(vmp)
+PUBLIC void map_printmap(vmp)
 struct vmproc *vmp;
 {
 	struct vir_region *vr;
@@ -60,14 +58,15 @@ struct vmproc *vmp;
 	for(vr = vmp->vm_regions; vr; vr = vr->next) {
 		struct phys_region *ph;
 		int nph = 0;
+		printf("map_printmap: map_name: %s\n", map_name(vr));
 		printf("\t0x%lx - 0x%lx (len 0x%lx), %s\n",
 			vr->vaddr, vr->vaddr + vr->length, vr->length,
-			vr->vaddr + vr->length, map_name(vr));
+			map_name(vr));
 		printf("\t\tphysical: ");
 		for(ph = vr->first; ph; ph = ph->next) {
 			printf("0x%lx-0x%lx (refs %d): phys 0x%lx ",
-				vr->vaddr + ph->ph->offset,
-				vr->vaddr + ph->ph->offset + ph->ph->length,
+				vr->vaddr + ph->offset,
+				vr->vaddr + ph->offset + ph->ph->length,
 				ph->ph->refcount,
 				ph->ph->phys);
 			nph++;
@@ -123,8 +122,8 @@ PUBLIC void map_sanitycheck(char *file, int line)
 			map_printmap(vmp);
 			printf("ph in vr 0x%lx: 0x%lx-0x%lx  refcount %d "
 				"but seencount %lu\n", 
-				vr, pr->ph->offset,
-				pr->ph->offset + pr->ph->length,
+				vr, pr->offset,
+				pr->offset + pr->ph->length,
 				pr->ph->refcount, pr->ph->seencount);
 		}
 		{
@@ -147,7 +146,7 @@ PUBLIC void map_sanitycheck(char *file, int line)
 			MYASSERT(pr->ph->refcount == n_others);
 		}
 		MYASSERT(pr->ph->refcount == pr->ph->seencount);
-		MYASSERT(!(pr->ph->offset % VM_PAGE_SIZE));
+		MYASSERT(!(pr->offset % VM_PAGE_SIZE));
 		MYASSERT(!(pr->ph->length % VM_PAGE_SIZE)););
 }
 #endif
@@ -156,14 +155,15 @@ PUBLIC void map_sanitycheck(char *file, int line)
 /*=========================================================================*
  *				map_ph_writept				*
  *=========================================================================*/
-PUBLIC int map_ph_writept(struct vmproc *vmp, struct vir_region *vr,
-	struct phys_block *pb, int *ropages, int *rwpages)
+PRIVATE int map_ph_writept(struct vmproc *vmp, struct vir_region *vr,
+	struct phys_region *pr, int *ropages, int *rwpages)
 {
 	int rw;
+	struct phys_block *pb = pr->ph;
 
 	vm_assert(!(vr->vaddr % VM_PAGE_SIZE));
 	vm_assert(!(pb->length % VM_PAGE_SIZE));
-	vm_assert(!(pb->offset % VM_PAGE_SIZE));
+	vm_assert(!(pr->offset % VM_PAGE_SIZE));
 	vm_assert(pb->refcount > 0);
 
 	if((vr->flags & VR_WRITABLE)
@@ -183,7 +183,7 @@ PUBLIC int map_ph_writept(struct vmproc *vmp, struct vir_region *vr,
 	}
 #endif
 
-	if(pt_writemap(&vmp->vm_pt, vr->vaddr + pb->offset,
+	if(pt_writemap(&vmp->vm_pt, vr->vaddr + pr->offset,
 	  pb->phys, pb->length, PTF_PRESENT | PTF_USER | rw,
 		WMF_OVERWRITE) != OK) {
 	    printf("VM: map_writept: pt_writemap failed\n");
@@ -308,7 +308,7 @@ int mapflags;
 		}
 		vm_assert(newregion->first);
 		vm_assert(!newregion->first->next);
-		if(map_ph_writept(vmp, newregion, newregion->first->ph, NULL, NULL) != OK) {
+		if(map_ph_writept(vmp, newregion, newregion->first, NULL, NULL) != OK) {
 			printf("VM: map_region_writept failed\n");
 			SLABFREE(newregion);
 			return NULL;
@@ -409,7 +409,8 @@ void pb_unreferenced(struct vir_region *region, struct phys_region *pr)
 			SLABSANE(pb->firstregion);
 			SLABSANE(pb->firstregion->parent);
 			if(map_ph_writept(pb->firstregion->parent->parent,
-				pb->firstregion->parent, pb, NULL, NULL) != OK) {
+				pb->firstregion->parent, pb->firstregion,
+				NULL, NULL) != OK) {
 				vm_panic("pb_unreferenced: writept", NO_NUM);
 			}
 		}
@@ -545,6 +546,8 @@ struct phys_region *physhint;
 		u32_t af = PAF_CLEAR;
 		if(region->flags & VR_PHYS64K)
 			af |= PAF_ALIGN64K;
+		if(region->flags & VR_LOWER16MB)
+			af |= PAF_LOWER16MB;
 		if((mem_clicks = ALLOC_MEM(clicks, af)) == NO_MEM) {
 			SLABFREE(newpb);
 			SLABFREE(newphysr);
@@ -559,12 +562,12 @@ struct phys_region *physhint;
 	/* New physical block. */
 	newpb->phys = mem;
 	newpb->refcount = 1;
-	newpb->offset = offset;
 	newpb->length = length;
 	newpb->firstregion = newphysr;
 	SLABSANE(newpb->firstregion);
 
 	/* New physical region. */
+	newphysr->offset = offset;
 	newphysr->ph = newpb;
 	newphysr->parent = region;
 	newphysr->next_ph_list = NULL;	/* No other references to this block. */
@@ -573,7 +576,7 @@ struct phys_region *physhint;
 	vm_assert(!(length % VM_PAGE_SIZE));
 	vm_assert(!(newpb->length % VM_PAGE_SIZE));
 	SANITYCHECK(SCL_DETAIL);
-	if(map_ph_writept(vmp, region, newpb, NULL, NULL) != OK) {
+	if(map_ph_writept(vmp, region, newphysr, NULL, NULL) != OK) {
 		if(what_mem == MAP_NONE)
 			FREE_MEM(mem_clicks, clicks);
 		SLABFREE(newpb);
@@ -581,17 +584,17 @@ struct phys_region *physhint;
 		return ENOMEM;
 	}
 
-	if(!region->first || offset < region->first->ph->offset) {
+	if(!region->first || offset < region->first->offset) {
 		/* Special case: offset is before start. */
 		if(region->first) {
-			vm_assert(offset + length <= region->first->ph->offset);
+			vm_assert(offset + length <= region->first->offset);
 		}
 		newphysr->next = region->first;
 		region->first = newphysr;
 	} else {
 		struct phys_region *physr;
 		for(physr = physhint; physr; physr = physr->next) {
-			if(!physr->next || physr->next->ph->offset > offset) {
+			if(!physr->next || physr->next->offset > offset) {
 				newphysr->next = physr->next;
 				physr->next = newphysr;
 				break;
@@ -649,7 +652,6 @@ struct phys_region *ph;
 	SLABSANE(ph->ph);
 	vm_assert(ph->ph->refcount > 0);
 	newpb->length = ph->ph->length;
-	newpb->offset = ph->ph->offset;
 	newpb->refcount = 1;
 	newpb->phys = newmem;
 	newpb->firstregion = ph;
@@ -675,7 +677,7 @@ struct phys_region *ph;
 	/* Update pagetable with new address.
 	 * This will also make it writable.
 	 */
-	r = map_ph_writept(vmp, region, ph->ph, NULL, NULL);
+	r = map_ph_writept(vmp, region, ph, NULL, NULL);
 	if(r != OK)
 		vm_panic("map_copy_ph_block: map_ph_writept failed", r);
 
@@ -708,7 +710,7 @@ int write;
 	SANITYCHECK(SCL_FUNCTIONS);
 
 	for(ph = region->first; ph; ph = ph->next)
-		if(ph->ph->offset <= offset && offset < ph->ph->offset + ph->ph->length)
+		if(ph->offset <= offset && offset < ph->offset + ph->ph->length)
 			break;
 
 	if(ph) {
@@ -718,9 +720,10 @@ int write;
 		vm_assert(ph->ph->refcount > 0);
 
 		if(ph->ph->refcount == 1)
-			r = map_ph_writept(vmp, region, ph->ph, NULL, NULL);
-		else
+			r = map_ph_writept(vmp, region, ph, NULL, NULL);
+		else {
 			r = map_copy_ph_block(vmp, region, ph);
+		}
 	} else {
 		/* Pagefault in non-existing block. Map in new block. */
 #if 0
@@ -756,8 +759,8 @@ int write;
 #define FREE_RANGE_HERE(er1, er2) {					\
 	struct phys_region *r1 = (er1), *r2 = (er2);			\
 	vir_bytes start = offset, end = offset + length;		\
-	if(r1) { start = MAX(start, r1->ph->offset + r1->ph->length); }	\
-	if(r2) { end   = MIN(end, r2->ph->offset); }			\
+	if(r1) { start = MAX(start, r1->offset + r1->ph->length); }	\
+	if(r2) { end   = MIN(end, r2->offset); }			\
 	if(start < end) {						\
 		int r;							\
 		SANITYCHECK(SCL_DETAIL);				\
@@ -797,7 +800,7 @@ int write;
 			SANITYCHECK(SCL_DETAIL);
 		  } else {
 			SANITYCHECK(SCL_DETAIL);
-			if((r=map_ph_writept(vmp, region, physr->ph, NULL, NULL)) != OK) {
+			if((r=map_ph_writept(vmp, region, physr, NULL, NULL)) != OK) {
 				printf("VM: map_ph_writept failed\n");
 				return r;
 			}
@@ -870,6 +873,7 @@ PRIVATE struct vir_region *map_copy_region(struct vir_region *vr)
 		newph->ph = ph->ph;
 		newph->next_ph_list = NULL;
 		newph->parent = newvr;
+		newph->offset = ph->offset;
 		if(prevph) prevph->next = newph;
 		else newvr->first = newph;
 		prevph = newph;
@@ -895,7 +899,7 @@ PUBLIC int map_writept(struct vmproc *vmp)
 
 	for(vr = vmp->vm_regions; vr; vr = vr->next)
 		for(ph = vr->first; ph; ph = ph->next) {
-			map_ph_writept(vmp, vr, ph->ph, &ropages, &rwpages);
+			map_ph_writept(vmp, vr, ph, &ropages, &rwpages);
 		}
 
 	return OK;
@@ -912,6 +916,9 @@ struct vmproc *src;
 	dst->vm_regions = NULL;
 
 	SANITYCHECK(SCL_FUNCTIONS);
+
+	PT_SANE(&src->vm_pt);
+
 	for(vr = src->vm_regions; vr; vr = vr->next) {
 		struct vir_region *newvr;
 		struct phys_region *orig_ph, *new_ph;
@@ -958,8 +965,12 @@ struct vmproc *src;
 	}
 	SANITYCHECK(SCL_DETAIL);
 
+	PT_SANE(&src->vm_pt);
+
 	map_writept(src);
+	PT_SANE(&src->vm_pt);
 	map_writept(dst);
+	PT_SANE(&dst->vm_pt);
 
 	SANITYCHECK(SCL_FUNCTIONS);
 	return OK;

servers/vm/region.h

@@ -6,7 +6,6 @@ struct phys_block {
 #if SANITYCHECKS
 	u32_t			seencount;
 #endif
-	vir_bytes		offset;	/* offset from start of vir region */
 	vir_bytes		length;	/* no. of contiguous bytes */
 	phys_bytes		phys;	/* physical memory */
 	u8_t			refcount;	/* Refcount of these pages */
@@ -19,6 +18,7 @@ struct phys_region {
 	struct phys_region	*next;	/* next contiguous block */
 	struct phys_block	*ph;
 	struct vir_region	*parent; /* Region that owns this phys_region. */
+	vir_bytes		offset;	/* offset from start of vir region */
 
 	/* list of phys_regions that reference the same phys_block */
 	struct phys_region	*next_ph_list;	
@@ -38,6 +38,7 @@ struct vir_region {
 #define VR_WRITABLE	0x01	/* Process may write here. */
 #define VR_NOPF		0x02	/* May not generate page faults. */
 #define VR_PHYS64K	0x04	/* Physical memory must be 64k aligned. */
+#define VR_LOWER16MB	0x08
 
 /* Mapping type: */
 #define VR_ANON		0x10	/* Memory to be cleared and allocated */

servers/vm/sanitycheck.h

@@ -34,7 +34,7 @@
 	for(vmp = vmproc; vmp <= &vmproc[_NR_PROCS]; vmp++) { \
 		if((vmp->vm_flags & (VMF_INUSE | VMF_HASPT)) == \
 			(VMF_INUSE | VMF_HASPT)) { \
-			pt_sanitycheck(&vmp->vm_pt, __FILE__, __LINE__); \
+			PT_SANE(&vmp->vm_pt); \
 		} \
 	} \
 	map_sanitycheck(__FILE__, __LINE__); \

servers/vm/utility.c

@@ -108,6 +108,9 @@ struct mem_map *map_ptr;                        /* memory to remove */
   }
   if (memp >= &mem_chunks[NR_MEMS])
   {
+	printf("VM: looking for memory at 0x%x, length 0x%x\n",
+		CLICK2ABS(map_ptr[T].mem_phys),
+		CLICK2ABS(map_ptr[T].mem_len));
         vm_panic("reserve_proc_mem: can't find map in mem_chunks ",
                 map_ptr[T].mem_phys);
   }

servers/vm/vm.h

@@ -5,6 +5,7 @@
 #define PAF_CLEAR	0x01	/* Clear physical memory. */
 #define PAF_CONTIG	0x02	/* Physically contiguous. */
 #define PAF_ALIGN64K	0x04	/* Aligned to 64k boundary. */
+#define PAF_LOWER16MB	0x08
 
 /* special value for v in pt_allocmap */
 #define AM_AUTO         ((u32_t) -1)
@@ -14,7 +15,7 @@
 
 /* Compile in asserts and custom sanity checks at all? */
 #define SANITYCHECKS	0
-#define VMSTATS		1
+#define VMSTATS		0
 
 /* If so, this level: */
 #define SCL_NONE	0	/* No sanity checks - vm_assert()s only. */

tools/issue.install

@@ -1,8 +1,8 @@
 
-Welcome to MINIX 3.1.3.
+Welcome to MINIX 3.1.4.
 
-This is an interim release, not final release quality. It is intended as
-a prerelease for developers.
+This snapshot is an interim release, not final release quality. It is
+intended as a prerelease for developers.
 
 The system is now running and many commands work normally.  To use MINIX
 in a serious way, you need to install it to your hard disk, which you
@@ -16,7 +16,6 @@ to begin installing the many software packages available.  After you
 have installed the packages, type 'xdm' to start X Windows if you have
 installed it.
 
-It is strongly recommended that you print and read the 'setup.pdf' file
-on the CD-ROM before staring the installation. The file is also
-available at www.minix3.org/doc under the 'Installation' heading.
-
+Before you begin the installation process, it is strongly recommended
+that you print and carefully read the installation instructions
+available on the MINIX 3 website: http://www.minix3.org.

tools/package_sources.install

@@ -0,0 +1,133 @@
+CSSC-1.0.1
+PopTart-0.9.7
+Xaw3d-1.5
+ackpack
+antiword-0.35
+apache-1.3.37
+apr-0.9.12
+apr-util-0.9.12
+ascii-3.8
+atk-1.9.0
+audio-1.0.0
+autoconf-2.59
+autoconf-2.60
+automake-1.9
+avra-0.7
+bash-3.0
+bc-1.06
+bchunk-1.2.0
+bcrypt-1.1
+bison-2.1
+catdoc-0.94.2
+cpio-2.5
+ctags-5.5.4
+cvs-1.11.21
+dact-0.8.39
+diffstat-1.39
+diffutils-2.8.1
+dungeon-2.7.1
+ede-1.1
+efltk-2.0.6
+elvis
+exim-4.66
+fb
+flawfinder-1.26
+flex-2.5.4
+flip-1.19
+fltk-1.1.7
+fltk-2.0.0-5220
+gawk-3.1.4
+gdbm-1.8.3
+gettext-0.14
+glib-1.2.10
+glib-2.10.0
+gnupg-1.4.8
+grep-2.5.1a
+groff-1.19.2
+gv-3.5.8
+gzip-1.2.4
+gzip-beta-1.3.3
+hexcat-0.0.3
+irssi-0.8.10
+jikes-1.22
+john-1.7.2
+jpeg-6b
+jwm-1.8
+kermit-2.1.1
+lcrack-20040914
+less-3.8.5
+lesstif-0.93.96
+libiconv-1.9.1
+libmcrypt-2.5.7
+libpng-1.2.8
+libungif-4.1.3
+libxml2-2.6.28
+links-2.1pre26
+lpr
+lynx2-8-5
+lzo-1.08
+lzop-1.01
+m4-1.4.3
+make-3.80
+mdf2iso
+mtools-3.9.7
+nano-1.3.12
+nasm-0.98.38
+nawk
+ncftp-3.1.9
+ncsa_httpd_1.4
+ncurses-5.5
+neon-0.25.5
+nethack-3.4.3-2
+netpbm-10.26.30
+nomarch-1.3
+nrg2iso-0.2
+nvi-1.79
+openssh-4.3p2
+openssl-0.9.8a
+patch-2.5.4
+pce-0.1.7
+pcre-4.5
+pdksh-5.2.14
+picasm112c
+pine-4.64
+pkg-config-0.20
+prng-3.0.2
+pscan-1.1
+pstotext-1.9
+psutils
+python-2.4.3
+rcs-5.7
+readline-5.1
+rman-3.2
+robodoc-4.0.18
+rsync-2.6.7
+screen-4.0.2
+sed-3.02
+shhmsg-1.4.1
+simhv35-2
+sip-0.4
+slang-1.4.9
+slrn-0.9.8.1
+sqlite-3.3.17
+src2tex-2.12
+ssmtp-2.61
+subversion-1.4.0
+tcl8.4.14
+texinfo-4.7
+tiff-3.8.2
+tk8.4.14
+unrtf-0.19.3
+unzip-5.52
+vice-1.19
+vice-1.20
+vile-9.4
+vim-6.3
+wdiff-0.5
+webcpp-0.8.4
+wget-1.5.3
+whichman-2.4
+xv-3.10a
+zip-2.31
+zlib-1.2.3
+zsh-4.2.6

tools/packages.install

@@ -0,0 +1,120 @@
+CSSC-1.0.1
+ImageMagick-6.3.1
+MPlayer-1.0rc1
+PopTart-0.9.7
+X11R6.8.2
+Xaw3d-1.5
+ackpack
+antiword-0.35
+apache-1.3.37
+apr-0.9.12
+apr-util-0.9.12
+ascii-3.8
+atk-1.9.0
+audio-1.0.0
+autoconf-2.59
+autoconf-2.60
+automake-1.9
+avra-0.7
+bash-3.0
+bc-1.06
+bchunk-1.2.0
+bcrypt-1.1
+binutils-2.16.1
+bison-2.1
+catdoc-0.94.2
+cvs-1.11.21
+diffutils-2.8.1
+dungeon-2.7.1
+ede-1.1
+efltk-2.0.6
+elvis
+emacs-21.4
+exim-4.66
+fb
+flawfinder-1.26
+flex-2.5.4
+fltk-1.1.7
+fltk-2.0.0-5220
+gawk-3.1.4
+gcc-4.1.1-libs
+gcc-4.1.1
+gettext-0.14
+ghostscript-8.54-gpl
+ghostscript-fonts-std-8.11
+glib-1.2.10
+gnu-coreutils-5.2.1
+gnupg-1.4.8
+grep-2.5.1a
+groff-1.19.2
+gv-3.5.8
+gzip-1.2.4
+hexcat-0.0.3
+irssi-0.8.10
+jikes-1.22
+john-1.7.2
+jpeg-6b
+jwm-1.8
+kermit-2.1.1
+lcrack-20040914
+less-3.8.5
+lesstif-0.93.96
+libiconv-1.9.1
+libpng-1.2.8
+libxml2-2.6.28
+links-2.1pre26
+lpr
+lynx2-8-5
+m4-1.4.3
+make-3.80
+mdf2iso
+mtools-3.9.7
+nano-1.3.12
+nawk
+ncftp-3.1.9
+ncurses-5.5
+neon-0.25.5
+nethack-3.4.3-2
+nomarch-1.3
+nvi-1.79
+openssh-4.3p2
+openssl-0.9.8a
+patch-2.5.4
+pce-0.1.7
+pdksh-5.2.14
+perl-5.8.7
+php-5.2.1
+pine-4.64
+pkg-config-0.20
+psutils
+python-2.4.3
+rcs-5.7
+readline-5.1
+rman-3.2
+rsync-2.6.7
+screen-4.0.2
+sed-3.02
+shhmsg-1.4.1
+simhv35-2
+sip-0.4
+slang-1.4.9
+slrn-0.9.8.1
+sqlite-3.3.17
+ssmtp-2.61
+subversion-1.4.0
+tcl8.4.14
+texinfo-4.7
+tiff-3.8.2
+tk8.4.14
+unrtf-0.19.3
+unzip-5.52
+vice-1.20
+vile-9.4
+vim-6.3
+webcpp-0.8.4
+wget-1.5.3
+whichman-2.4
+xv-3.10a
+zip-2.31
+zlib-1.2.3
+zsh-4.2.6

tools/release.sh

@@ -3,6 +3,7 @@
 set -e
 
 XBIN=usr/xbin
+BRANCHNAME=src.beng-working.r4327
 SRC=src
 
 # size of /tmp during build
@@ -10,6 +11,10 @@ TMPKB=32000
 
 PACKAGEDIR=/usr/bigports/Packages
 PACKAGESOURCEDIR=/usr/bigports/Sources
+# List of packages included on installation media
+PACKAGELIST=packages.install
+# List of package source included on installation media
+PACKAGESOURCELIST=package_sources.install
 secs=`expr 32 '*' 64`
 export SHELL=/bin/sh
 
@@ -205,24 +210,37 @@ cp -rp /usr/lib $RELEASEDIR/usr
 cp -rp /bin/bigsh /bin/sh /bin/echo $RELEASEDIR/bin
 cp -rp /usr/bin/make /usr/bin/install /usr/bin/yacc /usr/bin/flex $RELEASEDIR/usr/bin
 
-if [ -d $PACKAGEDIR -a -d $PACKAGESOURCEDIR -a $PACKAGES -ne 0 ]
+if [ -d $PACKAGEDIR -a -d $PACKAGESOURCEDIR -a -f $PACKAGELIST -a -f $PACKAGESOURCELIST -a $PACKAGES -ne 0 ]
 then	echo " * Indexing packages"
 	bintotal=0
-	( cd $PACKAGEDIR
-	  for p in *.tar.bz2
+	( for p in `cat $PACKAGELIST`
 	  do	
-		p="`echo $p | sed 's/.tar.bz2//'`"
-		descr="../$p/.descr"
+		descr="$PACKAGEDIR/../$p/.descr"
 		if [ -f "$descr" ]
 		then	echo "$p|`cat $descr`"
 		fi
-	  done | tee List
+	  done | tee $RELEASEPACKAGE/List
 	)
 	echo " * Transfering $PACKAGEDIR to $RELEASEPACKAGE"
-	cp $PACKAGEDIR/* $RELEASEPACKAGE/
+        for p in `cat $PACKAGELIST`
+        do
+               if [ -f $PACKAGEDIR/$p.tar.bz2 ]
+               then
+		  cp $PACKAGEDIR/$p.tar.bz2 $RELEASEPACKAGE/
+               else
+                  echo "Can't copy $PACKAGEDIR/$p.tar.bz2. Missing."
+               fi
+        done
 	echo " * Transfering $PACKAGESOURCEDIR to $RELEASEPACKAGESOURCES"
-	cp $PACKAGESOURCEDIR/* $RELEASEPACKAGESOURCES/ || true
-
+        for p in `cat $PACKAGESOURCELIST`
+        do
+               if [ -f $PACKAGESOURCEDIR/$p.tar.bz2 ]
+               then
+	          cp $PACKAGESOURCEDIR/$p.tar.bz2 $RELEASEPACKAGESOURCES/
+               else
+                  echo "Can't copy $PACKAGESOURCEDIR/$p.tar.bz2. Missing."
+               fi
+        done
 fi
 
 # Make sure compilers and libraries are bin-owned
@@ -232,10 +250,10 @@ chmod -R u+w $RELEASEDIR/usr/lib
 if [ "$COPY" -ne 1 ]
 then
 	echo " * Doing new svn export"
-	REPO=https://gforge.cs.vu.nl/svn/minix/trunk/$SRC
+	REPO=https://gforge.cs.vu.nl/svn/minix/branches/$BRANCHNAME
 	REVISION="`svn info $USERNAME $SVNREV $REPO | grep '^Revision: ' | awk '{ print $2 }'`"
 	echo "Doing export of revision $REVISION from $REPO."
-	( cd $RELEASEDIR/usr && svn $USERNAME export -r$REVISION $REPO )
+	( cd $RELEASEDIR/usr && svn $USERNAME export -r$REVISION $REPO && mv $BRANCHNAME $SRC )
 	REVTAG=r$REVISION
 	echo "