John Baldwin fdce57a042 Add an EARLY_AP_STARTUP option to start APs earlier during boot.
Currently, Application Processors (non-boot CPUs) are started by
MD code at SI_SUB_CPU, but they are kept waiting in a "pen" until
SI_SUB_SMP at which point they are released to run kernel threads.
SI_SUB_SMP is one of the last SYSINIT levels, so APs don't enter
the scheduler and start running threads until fairly late in the
boot.

This change moves SI_SUB_SMP up to just before software interrupt
threads are created allowing the APs to start executing kernel
threads much sooner (before any devices are probed).  This allows
several initialization routines that need to perform initialization
on all CPUs to now perform that initialization in one step rather
than having to defer the AP initialization to a second SYSINIT run
at SI_SUB_SMP.  It also permits all CPUs to be available for
handling interrupts before any devices are probed.

This last feature fixes a problem on with interrupt vector exhaustion.
Specifically, in the old model all device interrupts were routed
onto the boot CPU during boot.  Later after the APs were released at
SI_SUB_SMP, interrupts were redistributed across all CPUs.

However, several drivers for multiqueue hardware allocate N interrupts
per CPU in the system.  In a system with many CPUs, just a few drivers
doing this could exhaust the available pool of interrupt vectors on
the boot CPU as each driver was allocating N * mp_ncpu vectors on the
boot CPU.  Now, drivers will allocate interrupts on their desired CPUs
during boot meaning that only N interrupts are allocated from the boot
CPU instead of N * mp_ncpu.

Some other bits of code can also be simplified as smp_started is
now true much earlier and will now always be true for these bits of
code.  This removes the need to treat the single-CPU boot environment
as a special case.

As a transition aid, the new behavior is available under a new kernel
option (EARLY_AP_STARTUP).  This will allow the option to be turned off
if need be during initial testing.  I plan to enable this on x86 by
default in a followup commit in the next few days and to have all
platforms moved over before 11.0.  Once the transition is complete,
the option will be removed along with the !EARLY_AP_STARTUP code.

These changes have only been tested on x86.  Other platform maintainers
are encouraged to port their architectures over as well.  The main
things to check for are any uses of smp_started in MD code that can be
simplified and SI_SUB_SMP SYSINITs in MD code that can be removed in
the EARLY_AP_STARTUP case (e.g. the interrupt shuffling).

PR:		kern/199321
Reviewed by:	markj, gnn, kib
Sponsored by:	Netflix
2016-05-14 18:22:52 +00:00

363 lines
8.5 KiB
C

/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License, Version 1.0 only
* (the "License"). You may not use this file except in compliance
* with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*
* $FreeBSD$
*
*/
/*
* Copyright 2005 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/types.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/kmem.h>
#include <sys/smp.h>
#include <sys/dtrace_impl.h>
#include <sys/dtrace_bsd.h>
#include <machine/clock.h>
#include <machine/frame.h>
#include <machine/trap.h>
#include <vm/pmap.h>
#define DELAYBRANCH(x) ((int)(x) < 0)
extern dtrace_id_t dtrace_probeid_error;
extern int (*dtrace_invop_jump_addr)(struct trapframe *);
extern void dtrace_getnanotime(struct timespec *tsp);
int dtrace_invop(uintptr_t, struct trapframe *, uintptr_t);
void dtrace_invop_init(void);
void dtrace_invop_uninit(void);
typedef struct dtrace_invop_hdlr {
int (*dtih_func)(uintptr_t, struct trapframe *, uintptr_t);
struct dtrace_invop_hdlr *dtih_next;
} dtrace_invop_hdlr_t;
dtrace_invop_hdlr_t *dtrace_invop_hdlr;
int
dtrace_invop(uintptr_t addr, struct trapframe *frame, uintptr_t arg0)
{
dtrace_invop_hdlr_t *hdlr;
int rval;
for (hdlr = dtrace_invop_hdlr; hdlr != NULL; hdlr = hdlr->dtih_next)
if ((rval = hdlr->dtih_func(addr, frame, arg0)) != 0)
return (rval);
return (0);
}
void
dtrace_invop_add(int (*func)(uintptr_t, struct trapframe *, uintptr_t))
{
dtrace_invop_hdlr_t *hdlr;
hdlr = kmem_alloc(sizeof (dtrace_invop_hdlr_t), KM_SLEEP);
hdlr->dtih_func = func;
hdlr->dtih_next = dtrace_invop_hdlr;
dtrace_invop_hdlr = hdlr;
}
void
dtrace_invop_remove(int (*func)(uintptr_t, struct trapframe *, uintptr_t))
{
dtrace_invop_hdlr_t *hdlr = dtrace_invop_hdlr, *prev = NULL;
for (;;) {
if (hdlr == NULL)
panic("attempt to remove non-existent invop handler");
if (hdlr->dtih_func == func)
break;
prev = hdlr;
hdlr = hdlr->dtih_next;
}
if (prev == NULL) {
ASSERT(dtrace_invop_hdlr == hdlr);
dtrace_invop_hdlr = hdlr->dtih_next;
} else {
ASSERT(dtrace_invop_hdlr != hdlr);
prev->dtih_next = hdlr->dtih_next;
}
kmem_free(hdlr, 0);
}
/*ARGSUSED*/
void
dtrace_toxic_ranges(void (*func)(uintptr_t base, uintptr_t limit))
{
/*
* No toxic regions?
*/
}
void
dtrace_xcall(processorid_t cpu, dtrace_xcall_t func, void *arg)
{
cpuset_t cpus;
if (cpu == DTRACE_CPUALL)
cpus = all_cpus;
else
CPU_SETOF(cpu, &cpus);
smp_rendezvous_cpus(cpus, smp_no_rendevous_barrier, func,
smp_no_rendevous_barrier, arg);
}
static void
dtrace_sync_func(void)
{
}
void
dtrace_sync(void)
{
dtrace_xcall(DTRACE_CPUALL, (dtrace_xcall_t)dtrace_sync_func, NULL);
}
static int64_t tgt_cpu_tsc;
static int64_t hst_cpu_tsc;
static int64_t timebase_skew[MAXCPU];
static uint64_t nsec_scale;
/* See below for the explanation of this macro. */
/* This is taken from the amd64 dtrace_subr, to provide a synchronized timer
* between multiple processors in dtrace. Since PowerPC Timebases can be much
* lower than x86, the scale shift is 26 instead of 28, allowing for a 15.63MHz
* timebase.
*/
#define SCALE_SHIFT 26
static void
dtrace_gethrtime_init_cpu(void *arg)
{
uintptr_t cpu = (uintptr_t) arg;
if (cpu == curcpu)
tgt_cpu_tsc = mftb();
else
hst_cpu_tsc = mftb();
}
static void
dtrace_gethrtime_init(void *arg)
{
struct pcpu *pc;
uint64_t tb_f;
cpuset_t map;
int i;
tb_f = cpu_tickrate();
/*
* The following line checks that nsec_scale calculated below
* doesn't overflow 32-bit unsigned integer, so that it can multiply
* another 32-bit integer without overflowing 64-bit.
* Thus minimum supported Timebase frequency is 15.63MHz.
*/
KASSERT(tb_f > (NANOSEC >> (32 - SCALE_SHIFT)), ("Timebase frequency is too low"));
/*
* We scale up NANOSEC/tb_f ratio to preserve as much precision
* as possible.
* 2^26 factor was chosen quite arbitrarily from practical
* considerations:
* - it supports TSC frequencies as low as 15.63MHz (see above);
*/
nsec_scale = ((uint64_t)NANOSEC << SCALE_SHIFT) / tb_f;
/* The current CPU is the reference one. */
sched_pin();
timebase_skew[curcpu] = 0;
CPU_FOREACH(i) {
if (i == curcpu)
continue;
pc = pcpu_find(i);
CPU_SETOF(PCPU_GET(cpuid), &map);
CPU_SET(pc->pc_cpuid, &map);
smp_rendezvous_cpus(map, NULL,
dtrace_gethrtime_init_cpu,
smp_no_rendevous_barrier, (void *)(uintptr_t) i);
timebase_skew[i] = tgt_cpu_tsc - hst_cpu_tsc;
}
sched_unpin();
}
#ifdef EARLY_AP_STARTUP
SYSINIT(dtrace_gethrtime_init, SI_SUB_DTRACE, SI_ORDER_ANY,
dtrace_gethrtime_init, NULL);
#else
SYSINIT(dtrace_gethrtime_init, SI_SUB_SMP, SI_ORDER_ANY, dtrace_gethrtime_init,
NULL);
#endif
/*
* DTrace needs a high resolution time function which can
* be called from a probe context and guaranteed not to have
* instrumented with probes itself.
*
* Returns nanoseconds since boot.
*/
uint64_t
dtrace_gethrtime()
{
uint64_t timebase;
uint32_t lo;
uint32_t hi;
/*
* We split timebase value into lower and higher 32-bit halves and separately
* scale them with nsec_scale, then we scale them down by 2^28
* (see nsec_scale calculations) taking into account 32-bit shift of
* the higher half and finally add.
*/
timebase = mftb() - timebase_skew[curcpu];
lo = timebase;
hi = timebase >> 32;
return (((lo * nsec_scale) >> SCALE_SHIFT) +
((hi * nsec_scale) << (32 - SCALE_SHIFT)));
}
uint64_t
dtrace_gethrestime(void)
{
struct timespec curtime;
dtrace_getnanotime(&curtime);
return (curtime.tv_sec * 1000000000UL + curtime.tv_nsec);
}
/* Function to handle DTrace traps during probes. See powerpc/powerpc/trap.c */
int
dtrace_trap(struct trapframe *frame, u_int type)
{
/*
* A trap can occur while DTrace executes a probe. Before
* executing the probe, DTrace blocks re-scheduling and sets
* a flag in its per-cpu flags to indicate that it doesn't
* want to fault. On returning from the probe, the no-fault
* flag is cleared and finally re-scheduling is enabled.
*
* Check if DTrace has enabled 'no-fault' mode:
*/
if ((cpu_core[curcpu].cpuc_dtrace_flags & CPU_DTRACE_NOFAULT) != 0) {
/*
* There are only a couple of trap types that are expected.
* All the rest will be handled in the usual way.
*/
switch (type) {
/* Page fault. */
case EXC_DSI:
case EXC_DSE:
/* Flag a bad address. */
cpu_core[curcpu].cpuc_dtrace_flags |= CPU_DTRACE_BADADDR;
cpu_core[curcpu].cpuc_dtrace_illval = frame->dar;
/*
* Offset the instruction pointer to the instruction
* following the one causing the fault.
*/
frame->srr0 += sizeof(int);
return (1);
case EXC_ISI:
case EXC_ISE:
/* Flag a bad address. */
cpu_core[curcpu].cpuc_dtrace_flags |= CPU_DTRACE_BADADDR;
cpu_core[curcpu].cpuc_dtrace_illval = frame->srr0;
/*
* Offset the instruction pointer to the instruction
* following the one causing the fault.
*/
frame->srr0 += sizeof(int);
return (1);
default:
/* Handle all other traps in the usual way. */
break;
}
}
/* Handle the trap in the usual way. */
return (0);
}
void
dtrace_probe_error(dtrace_state_t *state, dtrace_epid_t epid, int which,
int fault, int fltoffs, uintptr_t illval)
{
dtrace_probe(dtrace_probeid_error, (uint64_t)(uintptr_t)state,
(uintptr_t)epid,
(uintptr_t)which, (uintptr_t)fault, (uintptr_t)fltoffs);
}
static int
dtrace_invop_start(struct trapframe *frame)
{
switch (dtrace_invop(frame->srr0, frame, frame->fixreg[3])) {
case DTRACE_INVOP_JUMP:
break;
case DTRACE_INVOP_BCTR:
frame->srr0 = frame->ctr;
break;
case DTRACE_INVOP_BLR:
frame->srr0 = frame->lr;
break;
case DTRACE_INVOP_MFLR_R0:
frame->fixreg[0] = frame->lr;
frame->srr0 = frame->srr0 + 4;
break;
default:
return (-1);
}
return (0);
}
void dtrace_invop_init(void)
{
dtrace_invop_jump_addr = dtrace_invop_start;
}
void dtrace_invop_uninit(void)
{
dtrace_invop_jump_addr = 0;
}