diff --git a/kernel/arch/earm/protect.c b/kernel/arch/earm/protect.c
index fe6df33d9..75b3ef897 100644
--- a/kernel/arch/earm/protect.c
+++ b/kernel/arch/earm/protect.c
@@ -138,7 +138,8 @@ void arch_boot_proc(struct boot_image *ip, struct proc *rp)
 		/* callbacks for use in the kernel */
 		execi.copymem = libexec_copy_memcpy;
 		execi.clearmem = libexec_clear_memset;
-		execi.allocmem_prealloc = libexec_pg_alloc;
+		execi.allocmem_prealloc_cleared = libexec_pg_alloc;
+		execi.allocmem_prealloc_junk = libexec_pg_alloc;
 		execi.allocmem_ondemand = libexec_pg_alloc;
 		execi.clearproc = NULL;
 
diff --git a/kernel/arch/i386/protect.c b/kernel/arch/i386/protect.c
index 22acc3350..5a98a41ad 100644
--- a/kernel/arch/i386/protect.c
+++ b/kernel/arch/i386/protect.c
@@ -414,7 +414,8 @@ void arch_boot_proc(struct boot_image *ip, struct proc *rp)
 		/* callbacks for use in the kernel */
 		execi.copymem = libexec_copy_memcpy;
 		execi.clearmem = libexec_clear_memset;
-		execi.allocmem_prealloc = libexec_pg_alloc;
+		execi.allocmem_prealloc_junk = libexec_pg_alloc;
+		execi.allocmem_prealloc_cleared = libexec_pg_alloc;
 		execi.allocmem_ondemand = libexec_pg_alloc;
 		execi.clearproc = NULL;
 
diff --git a/lib/libexec/exec_elf.c b/lib/libexec/exec_elf.c
index 625b3f087..24bfed966 100644
--- a/lib/libexec/exec_elf.c
+++ b/lib/libexec/exec_elf.c
@@ -150,7 +150,8 @@ int libexec_load_elf(struct exec_info *execi)
 
 	assert(execi->copymem);
 	assert(execi->clearmem);
-	assert(execi->allocmem_prealloc);
+	assert(execi->allocmem_prealloc_cleared);
+	assert(execi->allocmem_prealloc_junk);
 	assert(execi->allocmem_ondemand);
 
 	for (i = 0; i < hdr->e_phnum; i++) {
@@ -167,51 +168,124 @@ int libexec_load_elf(struct exec_info *execi)
 	for (i = 0; i < hdr->e_phnum; i++) {
 		vir_bytes seg_membytes, page_offset, p_vaddr, vaddr;
 		vir_bytes chunk, vfileend, vmemend;
+		off_t foffset, fbytes;
 		Elf_Phdr *ph = &phdr[i];
+		int try_mmap = 1;
+		u16_t clearend = 0;
+		int pagechunk;
+		int mmap_prot = PROT_READ;
+
+		if(!(ph->p_flags & PF_R)) {
+			printf("libexec: warning: unreadable segment\n");
+		}
+
+		if(ph->p_flags & PF_W) {
+			mmap_prot |= PROT_WRITE;
+#if ELF_DEBUG
+			printf("libexec: adding PROT_WRITE\n");
+#endif
+		} else {
+#if ELF_DEBUG
+			printf("libexec: not adding PROT_WRITE\n");
+#endif
+		}
+
 		if (ph->p_type != PT_LOAD || ph->p_memsz == 0) continue;
+
+		if((ph->p_vaddr % PAGE_SIZE) != (ph->p_offset % PAGE_SIZE)) {
+			printf("libexec: unaligned ELF program?\n");
+			try_mmap = 0;
+		}
+
+		if(!execi->memmap) {
+			try_mmap = 0;
+		}
+
+		foffset = ph->p_offset;
+		fbytes = ph->p_filesz;
 		vaddr = p_vaddr = ph->p_vaddr + execi->load_offset;
 		seg_membytes = ph->p_memsz;
+
 		page_offset = vaddr % PAGE_SIZE;
 		vaddr -= page_offset;
+		foffset -= page_offset;
 		seg_membytes += page_offset;
+		fbytes += page_offset;
+		vfileend  = p_vaddr + ph->p_filesz;
+
+		/* if there's usable memory after the file end, we have
+		 * to tell VM to clear the memory part of the page when it's
+		 * mapped in
+		 */
+		if((pagechunk = (vfileend % PAGE_SIZE))
+			&& ph->p_filesz < ph->p_memsz) {
+			clearend = PAGE_SIZE - pagechunk;
+		}
+
 		seg_membytes = roundup(seg_membytes, PAGE_SIZE);
+		fbytes = roundup(fbytes, PAGE_SIZE);
+
 		if(first || startv > vaddr) startv = vaddr;
 		first = 0;
 
-		/* make us some memory */
-		if(execi->allocmem_prealloc(execi, vaddr, seg_membytes) != OK) {
-			if(execi->clearproc) execi->clearproc(execi);
-			return ENOMEM;
-		}
-
+		if(try_mmap && execi->memmap(execi, vaddr, fbytes, foffset, clearend, mmap_prot) == OK) {
 #if ELF_DEBUG
-		printf("mmapped 0x%lx-0x%lx\n", vaddr, vaddr+seg_membytes);
+			printf("libexec: mmap 0x%lx-0x%lx done, clearend 0x%x\n",
+				vaddr, vaddr+fbytes, clearend);
 #endif
 
-		/* Copy executable section into it */
-		if(execi->copymem(execi, ph->p_offset, p_vaddr, ph->p_filesz) != OK) {
-			if(execi->clearproc) execi->clearproc(execi);
-			return ENOMEM;
-		}
+			if(seg_membytes > fbytes) {
+				int rem_mem = seg_membytes - fbytes;;
+				vir_bytes remstart = vaddr + fbytes;
+				if(execi->allocmem_ondemand(execi,
+					remstart, rem_mem) != OK) {
+					printf("libexec: mmap extra mem failed\n");
+					return ENOMEM;
+				}
+#if ELF_DEBUG
+				else printf("libexec: allocated 0x%lx-0x%lx\n",
+
+					remstart, remstart+rem_mem);
+#endif
+			}
+		} else {
+			if(try_mmap) printf("libexec: mmap failed\n");
+
+			/* make us some memory */
+			if(execi->allocmem_prealloc_junk(execi, vaddr, seg_membytes) != OK) {
+				if(execi->clearproc) execi->clearproc(execi);
+				return ENOMEM;
+			}
 
 #if ELF_DEBUG
-		printf("copied 0x%lx-0x%lx\n", p_vaddr, p_vaddr+ph->p_filesz);
+			printf("mmapped 0x%lx-0x%lx\n", vaddr, vaddr+seg_membytes);
 #endif
 
-		/* Clear remaining bits */
-		vfileend  = p_vaddr + ph->p_filesz;
-		vmemend = vaddr + seg_membytes;
-		if((chunk = p_vaddr - vaddr) > 0) {
+			/* Copy executable section into it */
+			if(execi->copymem(execi, ph->p_offset, p_vaddr, ph->p_filesz) != OK) {
+				if(execi->clearproc) execi->clearproc(execi);
+				return ENOMEM;
+			}
+
 #if ELF_DEBUG
-			printf("start clearing 0x%lx-0x%lx\n", vaddr, vaddr+chunk);
+			printf("copied 0x%lx-0x%lx\n", p_vaddr, p_vaddr+ph->p_filesz);
 #endif
-			execi->clearmem(execi, vaddr, chunk);
-		}
-		if((chunk = vmemend - vfileend) > 0) {
+
+			/* Clear remaining bits */
+			vmemend = vaddr + seg_membytes;
+			if((chunk = p_vaddr - vaddr) > 0) {
 #if ELF_DEBUG
-			printf("end clearing 0x%lx-0x%lx\n", vfileend, vaddr+chunk);
+				printf("start clearing 0x%lx-0x%lx\n", vaddr, vaddr+chunk);
 #endif
-			execi->clearmem(execi, vfileend, chunk);
+				execi->clearmem(execi, vaddr, chunk);
+			}
+	
+			if((chunk = vmemend - vfileend) > 0) {
+#if ELF_DEBUG
+				printf("end clearing 0x%lx-0x%lx\n", vfileend, vaddr+chunk);
+#endif
+				execi->clearmem(execi, vfileend, chunk);
+			}
 		}
 	}
 
diff --git a/lib/libexec/exec_general.c b/lib/libexec/exec_general.c
index 2ac942df9..a1afecfa7 100644
--- a/lib/libexec/exec_general.c
+++ b/lib/libexec/exec_general.c
@@ -18,7 +18,7 @@
 #include <sys/mman.h>
 #include <machine/elf.h>
 
-int libexec_alloc_mmap_prealloc(struct exec_info *execi, off_t vaddr, size_t len)
+int libexec_alloc_mmap_prealloc_junk(struct exec_info *execi, off_t vaddr, size_t len)
 {
 	if(minix_mmap_for(execi->proc_e, (void *) vaddr, len,
 		PROT_READ|PROT_WRITE|PROT_EXEC,
@@ -29,6 +29,17 @@ int libexec_alloc_mmap_prealloc(struct exec_info *execi, off_t vaddr, size_t len
 	return OK;
 }
 
+int libexec_alloc_mmap_prealloc_cleared(struct exec_info *execi, off_t vaddr, size_t len)
+{
+	if(minix_mmap_for(execi->proc_e, (void *) vaddr, len,
+		PROT_READ|PROT_WRITE|PROT_EXEC,
+		MAP_ANON|MAP_PREALLOC|MAP_FIXED, -1, 0) == MAP_FAILED) {
+		return ENOMEM;
+	}
+
+	return OK;
+}
+
 int libexec_alloc_mmap_ondemand(struct exec_info *execi, off_t vaddr, size_t len)
 {
 	if(minix_mmap_for(execi->proc_e, (void *) vaddr, len,
diff --git a/lib/libexec/libexec.h b/lib/libexec/libexec.h
index 8b6db7618..8597e74c3 100644
--- a/lib/libexec/libexec.h
+++ b/lib/libexec/libexec.h
@@ -16,6 +16,10 @@ typedef int (*libexec_allocfunc_t)(struct exec_info *execi,
 
 typedef int (*libexec_procclearfunc_t)(struct exec_info *execi);
 
+typedef int (*libexec_mmap_t)(struct exec_info *execi,
+	vir_bytes vaddr, vir_bytes len, vir_bytes foffset, u16_t clearend,
+	int protflags);
+
 struct exec_info {
     /* Filled in by libexec caller */
     endpoint_t  proc_e;                 /* Process endpoint */
@@ -33,9 +37,11 @@ struct exec_info {
     /* Callback pointers for use by libexec */
     libexec_loadfunc_t copymem;		/* Copy callback */
     libexec_clearfunc_t clearmem;	/* Clear callback */
-    libexec_allocfunc_t allocmem_prealloc; /* Alloc callback */
+    libexec_allocfunc_t allocmem_prealloc_cleared; /* Alloc callback */
+    libexec_allocfunc_t allocmem_prealloc_junk; /* Alloc callback */
     libexec_allocfunc_t allocmem_ondemand; /* Alloc callback */
     libexec_procclearfunc_t clearproc;	/* Clear process callback */
+    libexec_mmap_t memmap;		/* mmap callback */
     void *opaque;			/* Callback data */
 
     /* Filled in by libexec load function */
@@ -55,7 +61,8 @@ int libexec_load_elf(struct exec_info *execi);
 
 int libexec_copy_memcpy(struct exec_info *execi, off_t offset, off_t vaddr, size_t len);
 int libexec_clear_memset(struct exec_info *execi, off_t vaddr, size_t len);
-int libexec_alloc_mmap_prealloc(struct exec_info *execi, off_t vaddr, size_t len);
+int libexec_alloc_mmap_prealloc_cleared(struct exec_info *execi, off_t vaddr, size_t len);
+int libexec_alloc_mmap_prealloc_junk(struct exec_info *execi, off_t vaddr, size_t len);
 int libexec_alloc_mmap_ondemand(struct exec_info *execi, off_t vaddr, size_t len);
 int libexec_clearproc_vm_procctl(struct exec_info *execi);
 int libexec_clear_sys_memset(struct exec_info *execi, off_t vaddr, size_t len);
diff --git a/servers/rs/exec.c b/servers/rs/exec.c
index d0c950f67..ffbab2c78 100644
--- a/servers/rs/exec.c
+++ b/servers/rs/exec.c
@@ -133,7 +133,8 @@ static int do_exec(int proc_e, char *exec, size_t exec_len, char *progname,
 	execi.copymem = read_seg;
 	execi.clearproc = libexec_clearproc_vm_procctl;
 	execi.clearmem = libexec_clear_sys_memset;
-	execi.allocmem_prealloc = libexec_alloc_mmap_prealloc;
+	execi.allocmem_prealloc_cleared = libexec_alloc_mmap_prealloc_cleared;
+	execi.allocmem_prealloc_junk = libexec_alloc_mmap_prealloc_junk;
 	execi.allocmem_ondemand = libexec_alloc_mmap_ondemand;
 
 	for(i = 0; exec_loaders[i].load_object != NULL; i++) {
diff --git a/servers/vfs/exec.c b/servers/vfs/exec.c
index daad6a2fb..ce00e56c8 100644
--- a/servers/vfs/exec.c
+++ b/servers/vfs/exec.c
@@ -306,7 +306,8 @@ int pm_exec(endpoint_t proc_e, vir_bytes path, size_t path_len,
   execi.args.copymem = read_seg;
   execi.args.clearproc = libexec_clearproc_vm_procctl;
   execi.args.clearmem = libexec_clear_sys_memset;
-  execi.args.allocmem_prealloc = libexec_alloc_mmap_prealloc;
+  execi.args.allocmem_prealloc_cleared = libexec_alloc_mmap_prealloc_cleared;
+  execi.args.allocmem_prealloc_junk = libexec_alloc_mmap_prealloc_junk;
   execi.args.allocmem_ondemand = libexec_alloc_mmap_ondemand;
   execi.args.opaque = &execi;
 
diff --git a/servers/vm/main.c b/servers/vm/main.c
index 9cd14b444..caf148de5 100644
--- a/servers/vm/main.c
+++ b/servers/vm/main.c
@@ -287,7 +287,8 @@ void exec_bootproc(struct vmproc *vmp, struct boot_image *ip)
         execi->copymem = libexec_copy_physcopy;
         execi->clearproc = NULL;
         execi->clearmem = libexec_clear_sys_memset;
-        execi->allocmem_prealloc = libexec_alloc_vm_prealloc;
+	execi->allocmem_prealloc_junk = libexec_alloc_vm_prealloc;
+	execi->allocmem_prealloc_cleared = libexec_alloc_vm_prealloc;
         execi->allocmem_ondemand = libexec_alloc_vm_ondemand;
 
 	if(libexec_load_elf(execi) != OK)