From owner-svn-src-all@FreeBSD.ORG Sat May 29 09:07:41 2010 Return-Path: Delivered-To: svn-src-all@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 478D01065675; Sat, 29 May 2010 09:07:41 +0000 (UTC) (envelope-from nyan@FreeBSD.org) Received: from svn.freebsd.org (unknown [IPv6:2001:4f8:fff6::2c]) by mx1.freebsd.org (Postfix) with ESMTP id 3500D8FC13; Sat, 29 May 2010 09:07:41 +0000 (UTC) Received: from svn.freebsd.org (localhost [127.0.0.1]) by svn.freebsd.org (8.14.3/8.14.3) with ESMTP id o4T97fCA066450; Sat, 29 May 2010 09:07:41 GMT (envelope-from nyan@svn.freebsd.org) Received: (from nyan@localhost) by svn.freebsd.org (8.14.3/8.14.3/Submit) id o4T97fNZ066448; Sat, 29 May 2010 09:07:41 GMT (envelope-from nyan@svn.freebsd.org) Message-Id: <201005290907.o4T97fNZ066448@svn.freebsd.org> From: Takahashi Yoshihiro Date: Sat, 29 May 2010 09:07:41 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org X-SVN-Group: head MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Cc: Subject: svn commit: r208638 - head/sys/pc98/pc98 X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 29 May 2010 09:07:41 -0000 Author: nyan Date: Sat May 29 09:07:40 2010 New Revision: 208638 URL: http://svn.freebsd.org/changeset/base/208638 Log: MFi386: revision 178471 - Add an integer argument to idle to indicate how likely we are to wake from idle over the next tick. - Add a new MD routine, cpu_wake_idle() to wakeup idle threads who are suspended in cpu specific states. This function can fail and cause the scheduler to fall back to another mechanism (ipi). - Implement support for mwait in cpu_idle() on i386/amd64 machines that support it. mwait is a higher performance way to synchronize cpus as compared to hlt & ipis. - Allow selecting the idle routine by name via sysctl machdep.idle. This replaces machdep.cpu_idle_hlt. Only idle routines supported by the current machine are permitted. Modified: head/sys/pc98/pc98/machdep.c Modified: head/sys/pc98/pc98/machdep.c ============================================================================== --- head/sys/pc98/pc98/machdep.c Sat May 29 09:07:25 2010 (r208637) +++ head/sys/pc98/pc98/machdep.c Sat May 29 09:07:40 2010 (r208638) @@ -1108,6 +1108,7 @@ cpu_est_clockrate(int cpu_id, uint64_t * return (0); } + /* * Shutdown the CPU as much as possible */ @@ -1118,70 +1119,177 @@ cpu_halt(void) __asm__ ("hlt"); } -/* - * Hook to idle the CPU when possible. In the SMP case we default to - * off because a halted cpu will not currently pick up a new thread in the - * run queue until the next timer tick. If turned on this will result in - * approximately a 4.2% loss in real time performance in buildworld tests - * (but improves user and sys times oddly enough), and saves approximately - * 5% in power consumption on an idle machine (tests w/2xCPU 1.1GHz P3). - * - * XXX we need to have a cpu mask of idle cpus and generate an IPI or - * otherwise generate some sort of interrupt to wake up cpus sitting in HLT. - * Then we can have our cake and eat it too. - * - * XXX I'm turning it on for SMP as well by default for now. It seems to - * help lock contention somewhat, and this is critical for HTT. -Peter - */ -static int cpu_idle_hlt = 1; -TUNABLE_INT("machdep.cpu_idle_hlt", &cpu_idle_hlt); -SYSCTL_INT(_machdep, OID_AUTO, cpu_idle_hlt, CTLFLAG_RW, - &cpu_idle_hlt, 0, "Idle loop HLT enable"); - static void -cpu_idle_default(void) +cpu_idle_hlt(int busy) { /* - * we must absolutely guarentee that hlt is the - * absolute next instruction after sti or we - * introduce a timing window. + * we must absolutely guarentee that hlt is the next instruction + * after sti or we introduce a timing window. */ - __asm __volatile("sti; hlt"); + disable_intr(); + if (sched_runnable()) + enable_intr(); + else + __asm __volatile("sti; hlt"); } -/* - * Note that we have to be careful here to avoid a race between checking - * sched_runnable() and actually halting. If we don't do this, we may waste - * the time between calling hlt and the next interrupt even though there - * is a runnable process. - */ +static void +cpu_idle_spin(int busy) +{ + return; +} + +void (*cpu_idle_fn)(int) = cpu_idle_hlt; + void cpu_idle(int busy) { - -#ifdef SMP +#if defined(SMP) if (mp_grab_cpu_hlt()) return; #endif + cpu_idle_fn(busy); +} + +/* + * mwait cpu power states. Lower 4 bits are sub-states. + */ +#define MWAIT_C0 0xf0 +#define MWAIT_C1 0x00 +#define MWAIT_C2 0x10 +#define MWAIT_C3 0x20 +#define MWAIT_C4 0x30 + +#define MWAIT_DISABLED 0x0 +#define MWAIT_WOKEN 0x1 +#define MWAIT_WAITING 0x2 + +static void +cpu_idle_mwait(int busy) +{ + int *mwait; - if (cpu_idle_hlt) { - disable_intr(); - if (sched_runnable()) - enable_intr(); - else - (*cpu_idle_hook)(); + mwait = (int *)PCPU_PTR(monitorbuf); + *mwait = MWAIT_WAITING; + if (sched_runnable()) + return; + cpu_monitor(mwait, 0, 0); + if (*mwait == MWAIT_WAITING) + cpu_mwait(0, MWAIT_C1); +} + +static void +cpu_idle_mwait_hlt(int busy) +{ + int *mwait; + + mwait = (int *)PCPU_PTR(monitorbuf); + if (busy == 0) { + *mwait = MWAIT_DISABLED; + cpu_idle_hlt(busy); + return; } + *mwait = MWAIT_WAITING; + if (sched_runnable()) + return; + cpu_monitor(mwait, 0, 0); + if (*mwait == MWAIT_WAITING) + cpu_mwait(0, MWAIT_C1); } int cpu_idle_wakeup(int cpu) { + struct pcpu *pcpu; + int *mwait; - return (0); + if (cpu_idle_fn == cpu_idle_spin) + return (1); + if (cpu_idle_fn != cpu_idle_mwait && cpu_idle_fn != cpu_idle_mwait_hlt) + return (0); + pcpu = pcpu_find(cpu); + mwait = (int *)pcpu->pc_monitorbuf; + /* + * This doesn't need to be atomic since missing the race will + * simply result in unnecessary IPIs. + */ + if (cpu_idle_fn == cpu_idle_mwait_hlt && *mwait == MWAIT_DISABLED) + return (0); + *mwait = MWAIT_WOKEN; + + return (1); +} + +/* + * Ordered by speed/power consumption. + */ +struct { + void *id_fn; + char *id_name; +} idle_tbl[] = { + { cpu_idle_spin, "spin" }, + { cpu_idle_mwait, "mwait" }, + { cpu_idle_mwait_hlt, "mwait_hlt" }, + { cpu_idle_hlt, "hlt" }, + { NULL, NULL } +}; + +static int +idle_sysctl_available(SYSCTL_HANDLER_ARGS) +{ + char *avail, *p; + int error; + int i; + + avail = malloc(256, M_TEMP, M_WAITOK); + p = avail; + for (i = 0; idle_tbl[i].id_name != NULL; i++) { + if (strstr(idle_tbl[i].id_name, "mwait") && + (cpu_feature2 & CPUID2_MON) == 0) + continue; + p += sprintf(p, "%s, ", idle_tbl[i].id_name); + } + error = sysctl_handle_string(oidp, avail, 0, req); + free(avail, M_TEMP); + return (error); } -/* Other subsystems (e.g., ACPI) can hook this later. */ -void (*cpu_idle_hook)(void) = cpu_idle_default; +static int +idle_sysctl(SYSCTL_HANDLER_ARGS) +{ + char buf[16]; + int error; + char *p; + int i; + + p = "unknown"; + for (i = 0; idle_tbl[i].id_name != NULL; i++) { + if (idle_tbl[i].id_fn == cpu_idle_fn) { + p = idle_tbl[i].id_name; + break; + } + } + strncpy(buf, p, sizeof(buf)); + error = sysctl_handle_string(oidp, buf, sizeof(buf), req); + if (error != 0 || req->newptr == NULL) + return (error); + for (i = 0; idle_tbl[i].id_name != NULL; i++) { + if (strstr(idle_tbl[i].id_name, "mwait") && + (cpu_feature2 & CPUID2_MON) == 0) + continue; + if (strcmp(idle_tbl[i].id_name, buf)) + continue; + cpu_idle_fn = idle_tbl[i].id_fn; + return (0); + } + return (EINVAL); +} + +SYSCTL_PROC(_machdep, OID_AUTO, idle_available, CTLTYPE_STRING | CTLFLAG_RD, + 0, 0, idle_sysctl_available, "A", "list of available idle functions"); + +SYSCTL_PROC(_machdep, OID_AUTO, idle, CTLTYPE_STRING | CTLFLAG_RW, 0, 0, + idle_sysctl, "A", "currently selected idle function"); /* * Reset registers to default values on exec.