Module Name:    src
Committed By:   jmcneill
Date:           Sun Dec 13 20:39:20 UTC 2020

Modified Files:
        src/sys/dev/acpi: files.acpi
Added Files:
        src/sys/dev/acpi: acpi_cppc.c

Log Message:
acpi: Add support for Collaborative Processor Performance Control (CPPC).

CPPC is the modern replacement for processor performance controls (P-states
and T-states). The _CPC method on a processor device contains information
about supported performance levels of the device.

This driver reports the supported performance levels and allows for
setting the desired level using the machdep.cpufreq.cpuN.* sysctls. Note
that a "performance level" value is not guaranteed to represent a CPU
frequency, so treat the numbers reported as such.


To generate a diff of this commit:
cvs rdiff -u -r0 -r1.1 src/sys/dev/acpi/acpi_cppc.c
cvs rdiff -u -r1.118 -r1.119 src/sys/dev/acpi/files.acpi

Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.

Modified files:

Index: src/sys/dev/acpi/files.acpi
diff -u src/sys/dev/acpi/files.acpi:1.118 src/sys/dev/acpi/files.acpi:1.119
--- src/sys/dev/acpi/files.acpi:1.118	Sun Dec 13 20:27:53 2020
+++ src/sys/dev/acpi/files.acpi	Sun Dec 13 20:39:20 2020
@@ -1,4 +1,4 @@
-#	$NetBSD: files.acpi,v 1.118 2020/12/13 20:27:53 jmcneill Exp $
+#	$NetBSD: files.acpi,v 1.119 2020/12/13 20:39:20 jmcneill Exp $
 
 include "dev/acpi/acpica/files.acpica"
 
@@ -111,6 +111,11 @@ device	acpipcc
 attach	acpipcc at acpisdtbus
 file	dev/acpi/acpi_pcc.c		acpipcc
 
+# ACPI Collaborative Processor Performance Control
+device	acpicppc: acpipcc
+attach	acpicppc at acpinodebus
+file	dev/acpi/acpi_cppc.c		acpicppc
+
 # Serial interface
 attach	com at acpinodebus with com_acpi
 file	dev/acpi/com_acpi.c		com_acpi

Added files:

Index: src/sys/dev/acpi/acpi_cppc.c
diff -u /dev/null src/sys/dev/acpi/acpi_cppc.c:1.1
--- /dev/null	Sun Dec 13 20:39:20 2020
+++ src/sys/dev/acpi/acpi_cppc.c	Sun Dec 13 20:39:20 2020
@@ -0,0 +1,440 @@
+/* $NetBSD: acpi_cppc.c,v 1.1 2020/12/13 20:39:20 jmcneill Exp $ */
+
+/*-
+ * Copyright (c) 2020 Jared McNeill <jmcne...@invisible.ca>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
+ * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * ACPI Collaborative Processor Performance Control support.
+ */
+
+#include <sys/cdefs.h>
+__KERNEL_RCSID(0, "$NetBSD: acpi_cppc.c,v 1.1 2020/12/13 20:39:20 jmcneill Exp $");
+
+#include <sys/param.h>
+#include <sys/bus.h>
+#include <sys/cpu.h>
+#include <sys/device.h>
+#include <sys/kmem.h>
+#include <sys/sysctl.h>
+
+#include <dev/acpi/acpireg.h>
+#include <dev/acpi/acpivar.h>
+#include <dev/acpi/acpi_pcc.h>
+
+#include <external/bsd/acpica/dist/include/amlresrc.h>
+
+/* _CPC package elements */
+typedef enum CPCPackageElement {
+	CPCNumEntries,
+	CPCRevision,
+	CPCHighestPerformance,
+	CPCNominalPerformance,
+	CPCLowestNonlinearPerformance,
+	CPCLowestPerformance,
+	CPCGuaranteedPerformanceReg,
+	CPCDesiredPerformanceReg,
+	CPCMinimumPerformanceReg,
+	CPCMaximumPerformanceReg,
+	CPCPerformanceReductionToleranceReg,
+	CPCTimeWindowReg,
+	CPCCounterWraparoundTime,
+	CPCReferencePerformanceCounterReg,
+	CPCDeliveredPerformanceCounterReg,
+	CPCPerformanceLimitedReg,
+	CPCCPPCEnableReg,
+	CPCAutonomousSelectionEnable,
+	CPCAutonomousActivityWindowReg,
+	CPCEnergyPerformancePreferenceReg,
+	CPCReferencePerformance,
+	CPCLowestFrequency,
+	CPCNominalFrequency,
+} CPCPackageElement;
+
+/* PCC command numbers */
+#define	CPPC_PCC_READ	0x00
+#define	CPPC_PCC_WRITE	0x01
+
+struct cppc_softc {
+	device_t		sc_dev;
+	struct cpu_info	*	sc_cpuinfo;
+	ACPI_HANDLE		sc_handle;
+	ACPI_OBJECT *		sc_cpc;
+	u_int			sc_ncpc;
+
+	char *			sc_available;
+	int			sc_node_target;
+	int			sc_node_current;
+	ACPI_INTEGER		sc_max_target;
+	ACPI_INTEGER		sc_min_target;
+};
+
+static int		cppc_match(device_t, cfdata_t, void *);
+static void		cppc_attach(device_t, device_t, void *);
+
+static ACPI_STATUS 	cppc_parse_cpc(struct cppc_softc *);
+static ACPI_STATUS 	cppc_cpufreq_init(struct cppc_softc *);
+static ACPI_STATUS	cppc_read(struct cppc_softc *, CPCPackageElement,
+				  ACPI_INTEGER *);
+static ACPI_STATUS	cppc_write(struct cppc_softc *, CPCPackageElement,
+				   ACPI_INTEGER);
+
+CFATTACH_DECL_NEW(acpicppc, sizeof(struct cppc_softc),
+    cppc_match, cppc_attach, NULL, NULL);
+
+static const char * const compatible[] = {
+	"ACPI0007",	/* ACPI Processor Device */
+	NULL
+};
+
+static int
+cppc_match(device_t parent, cfdata_t cf, void *aux)
+{
+	struct acpi_attach_args * const aa = aux;
+	ACPI_HANDLE handle;
+	ACPI_STATUS rv;
+
+	if (aa->aa_node->ad_type != ACPI_TYPE_DEVICE) {
+		return 0;
+	}
+
+	if (acpi_match_hid(aa->aa_node->ad_devinfo, compatible) == 0) {
+		return 0;
+	}
+
+	rv = AcpiGetHandle(aa->aa_node->ad_handle, "_CPC", &handle);
+	if (ACPI_FAILURE(rv)) {
+		return 0;
+	}
+
+	if (acpi_match_cpu_handle(aa->aa_node->ad_handle) == NULL) {
+		return 0;
+	}
+
+	/* When CPPC and P-states/T-states are both available, prefer CPPC */
+	return 20;
+}
+
+static void
+cppc_attach(device_t parent, device_t self, void *aux)
+{
+	struct cppc_softc * const sc = device_private(self);
+	struct acpi_attach_args * const aa = aux;
+	ACPI_HANDLE handle = aa->aa_node->ad_handle;
+	struct cpu_info *ci;
+	ACPI_STATUS rv;
+
+	ci = acpi_match_cpu_handle(handle);
+	KASSERT(ci != NULL);
+
+	aprint_naive("\n");
+	aprint_normal(": Processor Performance Control (%s)\n", cpu_name(ci));
+
+	sc->sc_dev = self;
+	sc->sc_cpuinfo = ci;
+	sc->sc_handle = handle;
+
+	rv = cppc_parse_cpc(sc);
+	if (ACPI_FAILURE(rv)) {
+		aprint_error_dev(self, "failed to parse CPC package: %s\n",
+		    AcpiFormatException(rv));
+		return;
+	}
+
+	cppc_cpufreq_init(sc);
+}
+
+/*
+ * cppc_parse_cpc --
+ *
+ *	Read and verify the contents of the _CPC package.
+ */
+static ACPI_STATUS
+cppc_parse_cpc(struct cppc_softc *sc)
+{
+	ACPI_BUFFER buf;
+	ACPI_STATUS rv;
+
+	buf.Pointer = NULL;
+	buf.Length = ACPI_ALLOCATE_BUFFER;
+	rv = AcpiEvaluateObjectTyped(sc->sc_handle, "_CPC", NULL, &buf,
+	    ACPI_TYPE_PACKAGE);
+	if (ACPI_FAILURE(rv)) {
+		return rv;
+	}
+
+	sc->sc_cpc = (ACPI_OBJECT *)buf.Pointer;
+	if (sc->sc_cpc->Package.Count == 0) {
+		return AE_NOT_EXIST;
+	}
+	if (sc->sc_cpc->Package.Elements[CPCNumEntries].Type !=
+	    ACPI_TYPE_INTEGER) {
+		return AE_TYPE;
+	}
+	sc->sc_ncpc =
+	    sc->sc_cpc->Package.Elements[CPCNumEntries].Integer.Value;
+
+	return AE_OK;
+}
+
+/*
+ * cppc_cpufreq_sysctl --
+ *
+ *	sysctl helper function for machdep.cpu.cpuN.{target,current}
+ *	nodes.
+ */
+static int
+cppc_cpufreq_sysctl(SYSCTLFN_ARGS)
+{
+	struct cppc_softc * const sc = rnode->sysctl_data;
+	struct sysctlnode node;
+	u_int fq, oldfq = 0;
+	ACPI_INTEGER val;
+	ACPI_STATUS rv;
+	int error;
+
+	node = *rnode;
+	node.sysctl_data = &fq;
+
+	if (rnode->sysctl_num == sc->sc_node_target) {
+		rv = cppc_read(sc, CPCDesiredPerformanceReg, &val);
+	} else {
+		/*
+		 * XXX We should measure the delivered performance and
+		 *     report it here. For now, just report the desired
+		 *     performance level.
+		 */
+		rv = cppc_read(sc, CPCDesiredPerformanceReg, &val);
+	}
+	if (ACPI_FAILURE(rv)) {
+		return EIO;
+	}
+	if (val > UINT32_MAX) {
+		return ERANGE;
+	}
+	fq = (u_int)val;
+
+	if (rnode->sysctl_num == sc->sc_node_target) {
+		oldfq = fq;
+	}
+
+	error = sysctl_lookup(SYSCTLFN_CALL(&node));
+	if (error != 0 || newp == NULL) {
+		return error;
+	}
+
+	if (fq == oldfq || rnode->sysctl_num != sc->sc_node_target) {
+		return 0;
+	}
+
+	if (fq < sc->sc_min_target || fq > sc->sc_max_target) {
+		return EINVAL;
+	}
+
+	rv = cppc_write(sc, CPCDesiredPerformanceReg, fq);
+	if (ACPI_FAILURE(rv)) {
+		return EIO;
+	}
+
+	return 0;
+}
+
+/*
+ * cppc_cpufreq_init --
+ *
+ *	Create sysctl machdep.cpu.cpuN.* sysctl tree.
+ */
+static ACPI_STATUS
+cppc_cpufreq_init(struct cppc_softc *sc)
+{
+	static CPCPackageElement perf_regs[4] = {
+		CPCHighestPerformance,
+		CPCNominalPerformance,
+		CPCLowestNonlinearPerformance,
+		CPCLowestPerformance
+	};
+	ACPI_INTEGER perf[4], last;
+	const struct sysctlnode *node, *cpunode;
+	struct sysctllog *log = NULL;
+	struct cpu_info *ci = sc->sc_cpuinfo;
+	ACPI_STATUS rv;
+	int error, i, n;
+
+	/*
+	 * Read highest, nominal, lowest nonlinear, and lowest performance
+	 * levels and advertise this list of performance levels in the
+	 * machdep.cpufreq.cpuN.available sysctl.
+	 */
+	sc->sc_available = kmem_zalloc(
+	    strlen("########## ") * __arraycount(perf_regs), KM_SLEEP);
+	last = 0;
+	for (i = 0, n = 0; i < __arraycount(perf_regs); i++) {
+		rv = cppc_read(sc, perf_regs[i], &perf[i]);
+		if (ACPI_FAILURE(rv)) {
+			return rv;
+		}
+		if (perf[i] != last) {
+			char buf[12];
+			snprintf(buf, sizeof(buf), n ? " %u" : "%u",
+			    (u_int)perf[i]);
+			strcat(sc->sc_available, buf);
+			last = perf[i];
+			n++;
+		}
+	}
+	sc->sc_max_target = perf[0];
+	sc->sc_min_target = perf[3];
+
+	error = sysctl_createv(&log, 0, NULL, &node,
+	    CTLFLAG_PERMANENT, CTLTYPE_NODE, "machdep", NULL,
+	    NULL, 0, NULL, 0, CTL_MACHDEP, CTL_EOL);
+	if (error != 0) {
+		goto sysctl_failed;
+	}
+
+	error = sysctl_createv(&log, 0, &node, &node,
+	    0, CTLTYPE_NODE, "cpufreq", NULL,
+	    NULL, 0, NULL, 0, CTL_CREATE, CTL_EOL);
+	if (error != 0) {
+		goto sysctl_failed;
+	}
+
+	error = sysctl_createv(&log, 0, &node, &cpunode,
+	    0, CTLTYPE_NODE, cpu_name(ci), NULL,
+	    NULL, 0, NULL, 0, CTL_CREATE, CTL_EOL);
+	if (error != 0) {
+		goto sysctl_failed;
+	}
+
+	error = sysctl_createv(&log, 0, &cpunode, &node,
+	    CTLFLAG_READWRITE, CTLTYPE_INT, "target", NULL,
+	    cppc_cpufreq_sysctl, 0, (void *)sc, 0,
+	    CTL_CREATE, CTL_EOL);
+	if (error != 0) {
+		goto sysctl_failed;
+	}
+	sc->sc_node_target = node->sysctl_num;
+
+	error = sysctl_createv(&log, 0, &cpunode, &node,
+	    CTLFLAG_READONLY, CTLTYPE_INT, "current", NULL,
+	    cppc_cpufreq_sysctl, 0, (void *)sc, 0,
+	    CTL_CREATE, CTL_EOL);
+	if (error != 0) {
+		goto sysctl_failed;
+	}
+	sc->sc_node_current = node->sysctl_num;
+
+	error = sysctl_createv(&log, 0, &cpunode, &node,
+	    CTLFLAG_READONLY, CTLTYPE_STRING, "available", NULL,
+	    NULL, 0, sc->sc_available, 0,
+	    CTL_CREATE, CTL_EOL);
+	if (error != 0) {
+		goto sysctl_failed;
+	}
+
+	return AE_OK;
+
+sysctl_failed:
+	aprint_error_dev(sc->sc_dev, "couldn't create sysctl nodes: %d\n",
+	    error);
+	sysctl_teardown(&log);
+
+	return AE_ERROR;
+}
+
+/*
+ * cppc_read --
+ *
+ *	Read a value from the CPC package that contains either an integer
+ *	or indirect register reference.
+ */
+static ACPI_STATUS
+cppc_read(struct cppc_softc *sc, CPCPackageElement index, ACPI_INTEGER *val)
+{
+	ACPI_OBJECT *obj;
+	ACPI_GENERIC_ADDRESS addr;
+	ACPI_STATUS rv;
+
+	if (index >= sc->sc_ncpc) {
+		return AE_NOT_EXIST;
+	}
+
+	obj = &sc->sc_cpc->Package.Elements[index];
+	switch (obj->Type) {
+	case ACPI_TYPE_INTEGER:
+		*val = obj->Integer.Value;
+		return AE_OK;
+
+	case ACPI_TYPE_BUFFER:
+		if (obj->Buffer.Length <
+		    sizeof(AML_RESOURCE_GENERIC_REGISTER)) {
+			return AE_TYPE;
+		}
+		memcpy(&addr, obj->Buffer.Pointer +
+		    sizeof(AML_RESOURCE_LARGE_HEADER), sizeof(addr));
+		if (addr.SpaceId == ACPI_ADR_SPACE_PLATFORM_COMM) {
+			rv = pcc_message(&addr, CPPC_PCC_READ, PCC_READ, val);
+		} else {
+			rv = AcpiRead(val, &addr);
+		}
+		return rv;
+
+	default:
+		return AE_SUPPORT;
+	}
+}
+
+/*
+ * cppc_write --
+ *
+ *	Write a value based on the CPC package to the specified register.
+ */
+static ACPI_STATUS
+cppc_write(struct cppc_softc *sc, CPCPackageElement index, ACPI_INTEGER val)
+{
+	ACPI_OBJECT *obj;
+	ACPI_GENERIC_ADDRESS addr;
+	ACPI_STATUS rv;
+
+	if (index >= sc->sc_ncpc) {
+		return AE_NOT_EXIST;
+	}
+
+	obj = &sc->sc_cpc->Package.Elements[index];
+	if (obj->Type != ACPI_TYPE_BUFFER ||
+	    obj->Buffer.Length < sizeof(AML_RESOURCE_GENERIC_REGISTER)) {
+		return AE_TYPE;
+	}
+
+	memcpy(&addr, obj->Buffer.Pointer +
+	    sizeof(AML_RESOURCE_LARGE_HEADER), sizeof(addr));
+	if (addr.SpaceId == ACPI_ADR_SPACE_PLATFORM_COMM) {
+		rv = pcc_message(&addr, CPPC_PCC_WRITE, PCC_WRITE, &val);
+	} else {
+		rv = AcpiWrite(val, &addr);
+	}
+
+	return rv;
+}

Reply via email to