Add support for the extended FPU states on amd64, both for native

64bit and 32bit ABIs.  As a side-effect, it enables AVX on capable
CPUs.

In particular:

- Query the CPU support for XSAVE, list of the supported extensions
  and the required size of FPU save area. The hw.use_xsave tunable is
  provided for disabling XSAVE, and hw.xsave_mask may be used to
  select the enabled extensions.

- Remove the FPU save area from PCB and dynamically allocate the
  (run-time sized) user save area on the top of the kernel stack,
  right above the PCB. Reorganize the thread0 PCB initialization to
  postpone it after BSP is queried for save area size.

- The dumppcb, stoppcbs and susppcbs now do not carry the FPU state as
  well. FPU state is only useful for suspend, where it is saved in
  dynamically allocated suspfpusave area.

- Use XSAVE and XRSTOR to save/restore FPU state, if supported and
  enabled.

- Define new mcontext_t flag _MC_HASFPXSTATE, indicating that
  mcontext_t has a valid pointer to out-of-struct extended FPU
  state. Signal handlers are supplied with stack-allocated fpu
  state. The sigreturn(2) and setcontext(2) syscall honour the flag,
  allowing the signal handlers to inspect and manipilate extended
  state in the interrupted context.

- The getcontext(2) never returns extended state, since there is no
  place in the fixed-sized mcontext_t to place variable-sized save
  area. And, since mcontext_t is embedded into ucontext_t, makes it
  impossible to fix in a reasonable way.  Instead of extending
  getcontext(2) syscall, provide a sysarch(2) facility to query
  extended FPU state.

- Add ptrace(2) support for getting and setting extended state; while
  there, implement missed PT_I386_{GET,SET}XMMREGS for 32bit binaries.

- Change fpu_kern KPI to not expose struct fpu_kern_ctx layout to
  consumers, making it opaque. Internally, struct fpu_kern_ctx now
  contains a space for the extended state. Convert in-kernel consumers
  of fpu_kern KPI both on i386 and amd64.

First version of the support for AVX was submitted by Tim Bird
<tim.bird am sony com> on behalf of Sony. This version was written
from scratch.

Tested by:	pho (previous version), Yamagi Burmeister <lists yamagi org>
MFC after:	1 month

1 files changed, 63 insertions, 9 deletions

diff --git a/sys/amd64/amd64/vm_machdep.c b/sys/amd64/amd64/vm_machdep.c
index d05880fc2f96..acb21889ae3a 100644
--- a/sys/amd64/amd64/vm_machdep.c
+++ b/sys/amd64/amd64/vm_machdep.c
@@ -90,6 +90,51 @@ static u_int	cpu_reset_proxyid;
 static volatile u_int	cpu_reset_proxy_active;
 #endif
 
+struct savefpu *
+get_pcb_user_save_td(struct thread *td)
+{
+	vm_offset_t p;
+
+	p = td->td_kstack + td->td_kstack_pages * PAGE_SIZE -
+	    cpu_max_ext_state_size;
+	KASSERT((p % 64) == 0, ("Unaligned pcb_user_save area"));
+	return ((struct savefpu *)p);
+}
+
+struct savefpu *
+get_pcb_user_save_pcb(struct pcb *pcb)
+{
+	vm_offset_t p;
+
+	p = (vm_offset_t)(pcb + 1);
+	return ((struct savefpu *)p);
+}
+
+struct pcb *
+get_pcb_td(struct thread *td)
+{
+	vm_offset_t p;
+
+	p = td->td_kstack + td->td_kstack_pages * PAGE_SIZE -
+	    cpu_max_ext_state_size - sizeof(struct pcb);
+	return ((struct pcb *)p);
+}
+
+void *
+alloc_fpusave(int flags)
+{
+	struct pcb *res;
+	struct savefpu_ymm *sf;
+
+	res = malloc(cpu_max_ext_state_size, M_DEVBUF, flags);
+	if (use_xsave) {
+		sf = (struct savefpu_ymm *)res;
+		bzero(&sf->sv_xstate.sx_hd, sizeof(sf->sv_xstate.sx_hd));
+		sf->sv_xstate.sx_hd.xstate_bv = xsave_mask;
+	}
+	return (res);
+}
+
 /*
  * Finish a fork operation, with process p2 nearly set up.
  * Copy and update the pcb, set up the stack so that the child
@@ -127,15 +172,16 @@ cpu_fork(td1, p2, td2, flags)
 	fpuexit(td1);
 
 	/* Point the pcb to the top of the stack */
-	pcb2 = (struct pcb *)(td2->td_kstack +
-	    td2->td_kstack_pages * PAGE_SIZE) - 1;
+	pcb2 = get_pcb_td(td2);
 	td2->td_pcb = pcb2;
 
 	/* Copy td1's pcb */
 	bcopy(td1->td_pcb, pcb2, sizeof(*pcb2));
 
 	/* Properly initialize pcb_save */
-	pcb2->pcb_save = &pcb2->pcb_user_save;
+	pcb2->pcb_save = get_pcb_user_save_pcb(pcb2);
+	bcopy(get_pcb_user_save_td(td1), get_pcb_user_save_pcb(pcb2),
+	    cpu_max_ext_state_size);
 
 	/* Point mdproc and then copy over td1's contents */
 	mdp2 = &p2->p_md;
@@ -310,11 +356,17 @@ cpu_thread_swapout(struct thread *td)
 void
 cpu_thread_alloc(struct thread *td)
 {
-
-	td->td_pcb = (struct pcb *)(td->td_kstack +
-	    td->td_kstack_pages * PAGE_SIZE) - 1;
-	td->td_frame = (struct trapframe *)td->td_pcb - 1;
-	td->td_pcb->pcb_save = &td->td_pcb->pcb_user_save;
+	struct pcb *pcb;
+	struct xstate_hdr *xhdr;
+
+	td->td_pcb = pcb = get_pcb_td(td);
+	td->td_frame = (struct trapframe *)pcb - 1;
+	pcb->pcb_save = get_pcb_user_save_pcb(pcb);
+	if (use_xsave) {
+		xhdr = (struct xstate_hdr *)(pcb->pcb_save + 1);
+		bzero(xhdr, sizeof(*xhdr));
+		xhdr->xstate_bv = xsave_mask;
+	}
 }
 
 void
@@ -387,7 +439,9 @@ cpu_set_upcall(struct thread *td, struct thread *td0)
 	 */
 	bcopy(td0->td_pcb, pcb2, sizeof(*pcb2));
 	clear_pcb_flags(pcb2, PCB_FPUINITDONE | PCB_USERFPUINITDONE);
-	pcb2->pcb_save = &pcb2->pcb_user_save;
+	pcb2->pcb_save = get_pcb_user_save_pcb(pcb2);
+	bcopy(get_pcb_user_save_td(td0), pcb2->pcb_save,
+	    cpu_max_ext_state_size);
 	set_pcb_flags(pcb2, PCB_FULL_IRET);
 
 	/*