amd64: Add md process flags and first P_MD_PTI flag.

PTI mode for the process pmap on exec is activated iff P_MD_PTI is set.

On exec, the existing vmspace can be reused only if pti mode of the
pmap matches the P_MD_PTI flag of the process.  Add MD
cpu_exec_vmspace_reuse() callback for exec_new_vmspace() which can
vetoed reuse of the existing vmspace.

MFC note: md_flags change struct proc KBI.

Reviewed by:	jhb, markj
Tested by:	pho
Sponsored by:	The FreeBSD Foundation
MFC after:	1 week
Differential revision:	https://reviews.freebsd.org/D19514

sys/amd64/amd64/pmap.c

@@ -2853,6 +2853,7 @@ pmap_unuse_pt(pmap_t pmap, vm_offset_t va, pd_entry_t ptepde,
 void
 pmap_pinit0(pmap_t pmap)
 {
+	struct proc *p;
 	int i;
 
 	PMAP_LOCK_INIT(pmap);
@@ -2871,6 +2872,12 @@ pmap_pinit0(pmap_t pmap)
 		pmap->pm_pcids[i].pm_gen = 1;
 	}
 	pmap_activate_boot(pmap);
+	if (pti) {
+		p = curproc;
+		PROC_LOCK(p);
+		p->p_md.md_flags |= P_MD_KPTI;
+		PROC_UNLOCK(p);
+	}
 
 	if ((cpu_stdext_feature2 & CPUID_STDEXT2_PKU) != 0) {
 		pmap_pkru_ranges_zone = uma_zcreate("pkru ranges",
@@ -2957,7 +2964,7 @@ pmap_pinit_type(pmap_t pmap, enum pmap_type pm_type, int flags)
 	if (pm_type == PT_X86) {
 		pmap->pm_cr3 = pml4phys;
 		pmap_pinit_pml4(pml4pg);
-		if (pti) {
+		if ((curproc->p_md.md_flags & P_MD_KPTI) != 0) {
 			pml4pgu = vm_page_alloc(NULL, 0, VM_ALLOC_NORMAL |
 			    VM_ALLOC_NOOBJ | VM_ALLOC_WIRED | VM_ALLOC_WAITOK);
 			pmap->pm_pml4u = (pml4_entry_t *)PHYS_TO_DMAP(

sys/amd64/amd64/vm_machdep.c

@@ -369,6 +369,14 @@ cpu_thread_free(struct thread *td)
 	cpu_thread_clean(td);
 }
 
+bool
+cpu_exec_vmspace_reuse(struct proc *p, vm_map_t map)
+{
+
+	return (((curproc->p_md.md_flags & P_MD_KPTI) != 0) ==
+	    (vm_map_pmap(map)->pm_ucr3 != PMAP_NO_CR3));
+}
+
 void
 cpu_set_syscall_retval(struct thread *td, int error)
 {

sys/amd64/include/proc.h

@@ -40,7 +40,8 @@
 
 /*
  * List of locks
- *	k - only accessed by curthread
+ *	c  - proc lock
+ *	k  - only accessed by curthread
  *	pp - pmap.c:invl_gen_mtx
  */
 
@@ -69,8 +70,11 @@ struct mdthread {
 struct mdproc {
 	struct proc_ldt *md_ldt;	/* (t) per-process ldt */
 	struct system_segment_descriptor md_ldt_sd;
+	u_int md_flags;			/* (c) md process flags P_MD */
 };
 
+#define	P_MD_KPTI		0x00000001	/* Enable KPTI on exec */
+
 #define	KINFO_PROC_SIZE 1088
 #define	KINFO_PROC32_SIZE 768
 

sys/arm/arm/vm_machdep.c

@@ -345,3 +345,10 @@ cpu_exit(struct thread *td)
 {
 }
 
+bool
+cpu_exec_vmspace_reuse(struct proc *p __unused, vm_map_t map __unused)
+{
+
+	return (true);
+}
+

sys/arm64/arm64/vm_machdep.c

@@ -279,6 +279,13 @@ cpu_exit(struct thread *td)
 {
 }
 
+bool
+cpu_exec_vmspace_reuse(struct proc *p __unused, vm_map_t map __unused)
+{
+
+	return (true);
+}
+
 void
 swi_vm(void *v)
 {

sys/i386/i386/vm_machdep.c

@@ -382,6 +382,13 @@ cpu_thread_free(struct thread *td)
 	cpu_thread_clean(td);
 }
 
+bool
+cpu_exec_vmspace_reuse(struct proc *p __unused, vm_map_t map __unused)
+{
+
+	return (true);
+}
+
 void
 cpu_set_syscall_retval(struct thread *td, int error)
 {

sys/kern/kern_exec.c

@@ -1100,7 +1100,8 @@ exec_new_vmspace(struct image_params *imgp, struct sysentvec *sv)
 	else
 		sv_minuser = MAX(sv->sv_minuser, PAGE_SIZE);
 	if (vmspace->vm_refcnt == 1 && vm_map_min(map) == sv_minuser &&
-	    vm_map_max(map) == sv->sv_maxuser) {
+	    vm_map_max(map) == sv->sv_maxuser &&
+	    cpu_exec_vmspace_reuse(p, map)) {
 		shmexit(vmspace);
 		pmap_remove_pages(vmspace_pmap(vmspace));
 		vm_map_remove(map, vm_map_min(map), vm_map_max(map));

sys/kern/kern_thread.c

@@ -94,7 +94,7 @@ _Static_assert(offsetof(struct proc, p_filemon) == 0x3d0,
     "struct proc KBI p_filemon");
 _Static_assert(offsetof(struct proc, p_comm) == 0x3e8,
     "struct proc KBI p_comm");
-_Static_assert(offsetof(struct proc, p_emuldata) == 0x4c0,
+_Static_assert(offsetof(struct proc, p_emuldata) == 0x4c8,
     "struct proc KBI p_emuldata");
 #endif
 #ifdef __i386__

sys/mips/mips/vm_machdep.c

@@ -453,6 +453,13 @@ cpu_set_upcall(struct thread *td, void (*entry)(void *), void *arg,
 	 */
 }
 
+bool
+cpu_exec_vmspace_reuse(struct proc *p __unused, vm_map_t map __unused)
+{
+
+	return (true);
+}
+
 /*
  * Software interrupt handler for queued VM system processing.
  */

sys/powerpc/powerpc/vm_machdep.c

@@ -249,3 +249,10 @@ cpu_thread_swapout(struct thread *td)
 
 }
 
+bool
+cpu_exec_vmspace_reuse(struct proc *p __unused, vm_map_t map __unused)
+{
+
+	return (true);
+}
+

sys/riscv/riscv/vm_machdep.c

@@ -264,6 +264,13 @@ cpu_exit(struct thread *td)
 {
 }
 
+bool
+cpu_exec_vmspace_reuse(struct proc *p __unused, vm_map_t map __unused)
+{
+
+	return (true);
+}
+
 void
 swi_vm(void *v)
 {

sys/sparc64/sparc64/vm_machdep.c

@@ -373,6 +373,13 @@ cpu_fork_kthread_handler(struct thread *td, void (*func)(void *), void *arg)
 	fp->fr_local[1] = (u_long)arg;
 }
 
+bool
+cpu_exec_vmspace_reuse(struct proc *p __unused, vm_map_t map __unused)
+{
+
+	return (true);
+}
+
 int
 is_physical_memory(vm_paddr_t addr)
 {

sys/sys/proc.h

@@ -1093,6 +1093,7 @@ void	userret(struct thread *, struct trapframe *);
 void	cpu_exit(struct thread *);
 void	exit1(struct thread *, int, int) __dead2;
 void	cpu_copy_thread(struct thread *td, struct thread *td0);
+bool	cpu_exec_vmspace_reuse(struct proc *p, struct vm_map *map);
 int	cpu_fetch_syscall_args(struct thread *td);
 void	cpu_fork(struct thread *, struct proc *, struct thread *, int);
 void	cpu_fork_kthread_handler(struct thread *, void (*)(void *), void *);