soc/intel/common/block: Add HECI driver

Add common driver that can send/receive HECI messages. This driver is
inspired by Linux kernel mei driver and somewhat based on Skylake's.
Currently it has been only tested on Apollolake.

BUG=b:35586975
BRANCH=reef
TEST=tested on Apollolake to send single messages and receive both
fragmented and non-fragmented versions.

Change-Id: Ie3772700270f4f333292b80d59f79555851780f7
Signed-off-by: Andrey Petrov <andrey.petrov@intel.com>
Reviewed-on: https://review.coreboot.org/18547
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Tested-by: build bot (Jenkins)

4 files changed, 536 insertions, 0 deletions

diff --git a/src/soc/intel/common/block/cse/Kconfig b/src/soc/intel/common/block/cse/Kconfig
new file mode 100644
index 0000000000..441ff93cd7
--- /dev/null
+++ b/src/soc/intel/common/block/cse/Kconfig
@@ -0,0 +1,7 @@
+config SOC_INTEL_COMMON_BLOCK_CSE
+	bool
+	default n
+	help
+	  Driver for communication with Converged Security Engine (CSE)
+	  over Host Embedded Controller Interface (HECI)
+
diff --git a/src/soc/intel/common/block/cse/Makefile.inc b/src/soc/intel/common/block/cse/Makefile.inc
new file mode 100644
index 0000000000..fe7b13c7bb
--- /dev/null
+++ b/src/soc/intel/common/block/cse/Makefile.inc
@@ -0,0 +1,2 @@
+romstage-$(CONFIG_SOC_INTEL_COMMON_BLOCK_CSE) += cse.c
+ramstage-$(CONFIG_SOC_INTEL_COMMON_BLOCK_CSE) += cse.c
diff --git a/src/soc/intel/common/block/cse/cse.c b/src/soc/intel/common/block/cse/cse.c
new file mode 100644
index 0000000000..77cedb1407
--- /dev/null
+++ b/src/soc/intel/common/block/cse/cse.c
@@ -0,0 +1,477 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright 2017 Intel Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include <arch/early_variables.h>
+#include <commonlib/helpers.h>
+#include <console/console.h>
+#include <delay.h>
+#include <device/pci.h>
+#include <device/pci_ids.h>
+#include <device/pci_ops.h>
+#include <intelblocks/cse.h>
+#include <soc/pci_devs.h>
+#include <soc/pci_ids.h>
+#include <string.h>
+#include <timer.h>
+
+/* default window for early boot, must be at least 12 bytes in size */
+#define HECI1_BASE_ADDRESS	0xfed1a000
+
+/* Wait up to 15 sec for HECI to get ready */
+#define HECI_DELAY_READY	(15 * 1000)
+/* Wait up to 100 usec between circullar buffer polls */
+#define HECI_DELAY		100
+/* Wait up to 5 sec for CSE to chew something we sent */
+#define HECI_SEND_TIMEOUT	(5 * 1000)
+/* Wait up to 5 sec for CSE to blurp a reply */
+#define HECI_READ_TIMEOUT	(5 * 1000)
+
+#define SLOT_SIZE		sizeof(uint32_t)
+
+#define MMIO_CSE_CB_WW		0x00
+#define MMIO_HOST_CSR		0x04
+#define MMIO_CSE_CB_RW		0x08
+#define MMIO_CSE_CSR		0x0c
+
+#define CSR_IE			(1 << 0)
+#define CSR_IS			(1 << 1)
+#define CSR_IG			(1 << 2)
+#define CSR_READY		(1 << 3)
+#define CSR_RESET		(1 << 4)
+#define CSR_RP_START		8
+#define CSR_RP			(((1 << 8) - 1) << CSR_RP_START)
+#define CSR_WP_START		16
+#define CSR_WP			(((1 << 8) - 1) << CSR_WP_START)
+#define CSR_CBD_START		24
+#define CSR_CBD			(((1 << 8) - 1) << CSR_CBD_START)
+
+#define MEI_HDR_IS_COMPLETE	(1 << 31)
+#define MEI_HDR_LENGTH_START	16
+#define MEI_HDR_LENGTH_SIZE	9
+#define MEI_HDR_LENGTH		(((1 << MEI_HDR_LENGTH_SIZE) - 1) \
+					<< MEI_HDR_LENGTH_START)
+#define MEI_HDR_HOST_ADDR_START	8
+#define MEI_HDR_HOST_ADDR	(((1 << 8) - 1) << MEI_HDR_HOST_ADDR_START)
+#define MEI_HDR_CSE_ADDR_START	0
+#define MEI_HDR_CSE_ADDR	(((1 << 8) - 1) << MEI_HDR_CSE_ADDR_START)
+
+
+struct cse_device {
+	uintptr_t sec_bar;
+} g_cse CAR_GLOBAL;
+
+/*
+ * Initialize the device with provided temporary BAR. If BAR is 0 use a
+ * default. This is intended for pre-mem usage only where BARs haven't been
+ * assigned yet and devices are not enabled.
+ */
+void heci_init(uintptr_t tempbar)
+{
+	struct cse_device *cse = car_get_var_ptr(&g_cse);
+	device_t dev = HECI1_DEV;
+	u8 pcireg;
+
+	/* Assume it is already initialized, nothing else to do */
+	if (cse->sec_bar)
+		return;
+
+	/* Use default pre-ram bar */
+	if (!tempbar)
+		tempbar = HECI1_BASE_ADDRESS;
+
+	/* Assign Resources to HECI1 */
+	/* Clear BIT 1-2 of Command Register */
+	pcireg = pci_read_config8(dev, PCI_COMMAND);
+	pcireg &= ~(PCI_COMMAND_MASTER | PCI_COMMAND_MEMORY);
+	pci_write_config8(dev, PCI_COMMAND, pcireg);
+
+	/* Program Temporary BAR for HECI1 */
+	pci_write_config32(dev, PCI_BASE_ADDRESS_0, tempbar);
+	pci_write_config32(dev, PCI_BASE_ADDRESS_1, 0x0);
+
+	/* Enable Bus Master and MMIO Space */
+	pcireg = pci_read_config8(dev, PCI_COMMAND);
+	pcireg |= PCI_COMMAND_MASTER | PCI_COMMAND_MEMORY;
+	pci_write_config8(dev, PCI_COMMAND, pcireg);
+
+	cse->sec_bar = tempbar;
+}
+
+static uint32_t read_bar(uint32_t offset)
+{
+	struct cse_device *cse = car_get_var_ptr(&g_cse);
+	return read32((void *)(cse->sec_bar + offset));
+}
+
+static void write_bar(uint32_t offset, uint32_t val)
+{
+	struct cse_device *cse = car_get_var_ptr(&g_cse);
+	return write32((void *)(cse->sec_bar + offset), val);
+}
+
+static uint32_t read_cse_csr(void)
+{
+	return read_bar(MMIO_CSE_CSR);
+}
+
+static uint32_t read_host_csr(void)
+{
+	return read_bar(MMIO_HOST_CSR);
+}
+
+static void write_host_csr(uint32_t data)
+{
+	write_bar(MMIO_HOST_CSR, data);
+}
+
+static size_t filled_slots(uint32_t data)
+{
+	uint8_t wp, rp;
+	rp = data >> CSR_RP_START;
+	wp = data >> CSR_WP_START;
+	return (uint8_t) (wp - rp);
+}
+
+static size_t cse_filled_slots(void)
+{
+	return filled_slots(read_cse_csr());
+}
+
+static size_t host_empty_slots(void)
+{
+	uint32_t csr;
+	csr = read_host_csr();
+
+	return ((csr & CSR_CBD) >> CSR_CBD_START) - filled_slots(csr);
+}
+
+static void clear_int(void)
+{
+	uint32_t csr;
+	csr = read_host_csr();
+	csr |= CSR_IS;
+	write_host_csr(csr);
+}
+
+static uint32_t read_slot(void)
+{
+	return read_bar(MMIO_CSE_CB_RW);
+}
+
+static void write_slot(uint32_t val)
+{
+	write_bar(MMIO_CSE_CB_WW, val);
+}
+
+static int wait_write_slots(size_t cnt)
+{
+	struct stopwatch sw;
+
+	stopwatch_init_msecs_expire(&sw, HECI_SEND_TIMEOUT);
+	while (host_empty_slots() < cnt) {
+		udelay(HECI_DELAY);
+		if (stopwatch_expired(&sw)) {
+			printk(BIOS_ERR, "HECI: timeout, buffer not drained\n");
+			return 0;
+		}
+	}
+	return 1;
+}
+
+static int wait_read_slots(size_t cnt)
+{
+	struct stopwatch sw;
+
+	stopwatch_init_msecs_expire(&sw, HECI_READ_TIMEOUT);
+	while (cse_filled_slots() < cnt) {
+		udelay(HECI_DELAY);
+		if (stopwatch_expired(&sw)) {
+			printk(BIOS_ERR, "HECI: timed out reading answer!\n");
+			return 0;
+		}
+	}
+	return 1;
+}
+
+/* get number of full 4-byte slots */
+static size_t bytes_to_slots(size_t bytes)
+{
+	return ALIGN_UP(bytes, SLOT_SIZE) / SLOT_SIZE;
+}
+
+static int cse_ready(void)
+{
+	uint32_t csr;
+	csr = read_cse_csr();
+	return csr & CSR_READY;
+}
+
+static int wait_heci_ready(void)
+{
+	struct stopwatch sw;
+
+	stopwatch_init_msecs_expire(&sw, HECI_DELAY_READY);
+	while (!cse_ready()) {
+		udelay(HECI_DELAY);
+		if (stopwatch_expired(&sw))
+			return 0;
+	}
+
+	return 1;
+}
+
+static void host_gen_interrupt(void)
+{
+	uint32_t csr;
+	csr = read_host_csr();
+	csr |= CSR_IG;
+	write_host_csr(csr);
+}
+
+static size_t hdr_get_length(uint32_t hdr)
+{
+	return (hdr & MEI_HDR_LENGTH) >> MEI_HDR_LENGTH_START;
+}
+
+static int
+send_one_message(uint32_t hdr, const void *buff)
+{
+	size_t pend_len, pend_slots, remainder, i;
+	uint32_t tmp;
+	const uint32_t *p = buff;
+
+	/* Get space for the header */
+	if (!wait_write_slots(1))
+		return 0;
+
+	/* First, write header */
+	write_slot(hdr);
+
+	pend_len = hdr_get_length(hdr);
+	pend_slots = bytes_to_slots(pend_len);
+
+	if (!wait_write_slots(pend_slots))
+		return 0;
+
+	/* Write the body in whole slots */
+	i = 0;
+	while (i < ALIGN_DOWN(pend_len, SLOT_SIZE)) {
+		write_slot(*p++);
+		i += SLOT_SIZE;
+	}
+
+	remainder = pend_len % SLOT_SIZE;
+	/* Pad to 4 bytes not touching caller's buffer */
+	if (remainder) {
+		memcpy(&tmp, p, remainder);
+		write_slot(tmp);
+	}
+
+	host_gen_interrupt();
+
+	/* Make sure nothing bad happened during transmission */
+	if (!cse_ready())
+		return 0;
+
+	return pend_len;
+}
+
+int
+heci_send(const void *msg, size_t len, uint8_t host_addr, uint8_t client_addr)
+{
+	uint32_t csr, hdr;
+	size_t sent = 0, remaining, cb_size, max_length;
+	uint8_t *p = (uint8_t *) msg;
+
+	if (!msg || !len)
+		return 0;
+
+	clear_int();
+
+	if (!wait_heci_ready()) {
+		printk(BIOS_ERR, "HECI: not ready\n");
+		return 0;
+	}
+
+	csr = read_cse_csr();
+	cb_size = ((csr & CSR_CBD) >> CSR_CBD_START) * SLOT_SIZE;
+	/*
+	 * Reserve one slot for the header. Limit max message length by 9
+	 * bits that are available in the header.
+	 */
+	max_length = MIN(cb_size, (1 << MEI_HDR_LENGTH_SIZE) - 1) - SLOT_SIZE;
+	remaining = len;
+
+	/*
+	 * Fragment the message into smaller messages not exceeding useful
+	 * circullar buffer length. Mark last message complete.
+	 */
+	do {
+		hdr = MIN(max_length, remaining) << MEI_HDR_LENGTH_START;
+		hdr |= client_addr << MEI_HDR_CSE_ADDR_START;
+		hdr |= host_addr << MEI_HDR_HOST_ADDR_START;
+		hdr |= (MIN(max_length, remaining) == remaining) ?
+						MEI_HDR_IS_COMPLETE: 0;
+		sent = send_one_message(hdr, p);
+		p += sent;
+		remaining -= sent;
+	} while (remaining > 0 && sent != 0);
+
+	return remaining == 0;
+}
+
+static size_t
+recv_one_message(uint32_t *hdr, void *buff, size_t maxlen)
+{
+	uint32_t reg, *p = buff;
+	size_t recv_slots, recv_len, remainder, i;
+
+	/* first get the header */
+	if (!wait_read_slots(1))
+		return 0;
+
+	*hdr = read_slot();
+	recv_len = hdr_get_length(*hdr);
+
+	if (!recv_len)
+		printk(BIOS_WARNING, "HECI: message is zero-sized\n");
+
+	recv_slots = bytes_to_slots(recv_len);
+
+	i = 0;
+	if (recv_len > maxlen) {
+		printk(BIOS_ERR, "HECI: response is too big\n");
+		return 0;
+	}
+
+	/* wait for the rest of messages to arrive */
+	wait_read_slots(recv_slots);
+
+	/* fetch whole slots first */
+	while (i < ALIGN_DOWN(recv_len, SLOT_SIZE)) {
+		*p++ = read_slot();
+		i += SLOT_SIZE;
+	}
+
+	remainder = recv_len % SLOT_SIZE;
+
+	if (remainder) {
+		reg = read_slot();
+		memcpy(p, &reg, remainder);
+	}
+
+	return recv_len;
+}
+
+int heci_receive(void *buff, size_t *maxlen)
+{
+	size_t left, received;
+	uint32_t hdr = 0;
+	uint8_t *p = buff;
+
+	if (!buff || !maxlen || !*maxlen)
+		return 0;
+
+	left = *maxlen;
+
+	clear_int();
+
+	if (!wait_heci_ready()) {
+		printk(BIOS_ERR, "HECI: not ready\n");
+		return 0;
+	}
+
+	/*
+	 * Receive multiple packets until we meet one marked complete or we run
+	 * out of space in caller-provided buffer.
+	 */
+	do {
+		received = recv_one_message(&hdr, p, left);
+		left -= received;
+		p += received;
+		/* If we read out everything ping to send more */
+		if (!(hdr & MEI_HDR_IS_COMPLETE) && !cse_filled_slots())
+			host_gen_interrupt();
+	} while (received && !(hdr & MEI_HDR_IS_COMPLETE) && left > 0);
+
+	*maxlen = p - (uint8_t *) buff;
+
+	/* If ME is not ready, something went wrong and we received junk */
+	if (!cse_ready())
+		return 0;
+
+	return !!((hdr & MEI_HDR_IS_COMPLETE) && received);
+}
+
+/*
+ * Attempt to reset the device. This is useful when host and ME are out
+ * of sync during transmission or ME didn't understand the message.
+ */
+int heci_reset(void)
+{
+	uint32_t csr;
+
+	/* Send reset request */
+	csr = read_host_csr();
+	csr |= CSR_RESET;
+	csr |= CSR_IG;
+	write_host_csr(csr);
+
+	if (wait_heci_ready()) {
+		/* Device is back on its imaginary feet, clear reset */
+		csr = read_host_csr();
+		csr &= ~CSR_RESET;
+		csr |= CSR_IG;
+		csr |= CSR_READY;
+		write_host_csr(csr);
+		return 1;
+	}
+
+	printk(BIOS_CRIT, "HECI: reset failed\n");
+
+	return 0;
+}
+
+#if ENV_RAMSTAGE
+
+static void update_sec_bar(struct device *dev)
+{
+	g_cse.sec_bar = find_resource(dev, PCI_BASE_ADDRESS_0)->base;
+}
+
+static void cse_set_resources(struct device *dev)
+{
+	if (dev->path.pci.devfn == HECI1_DEVFN)
+		update_sec_bar(dev);
+
+	pci_dev_set_resources(dev);
+}
+
+static struct device_operations cse_ops = {
+	.set_resources		= cse_set_resources,
+	.read_resources		= pci_dev_read_resources,
+	.enable_resources	= pci_dev_enable_resources,
+	.init			= pci_dev_init,
+	.enable_resources	= pci_dev_enable_resources
+};
+
+static const struct pci_driver cse_driver __pci_driver = {
+	.ops			= &cse_ops,
+	.vendor			= PCI_VENDOR_ID_INTEL,
+	/* SoC/chipset needs to provide PCI device ID */
+	.device			= PCI_DEVICE_ID_HECI1
+};
+
+#endif
diff --git a/src/soc/intel/common/block/include/intelblocks/cse.h b/src/soc/intel/common/block/include/intelblocks/cse.h
new file mode 100644
index 0000000000..d7c4d9f93e
--- /dev/null
+++ b/src/soc/intel/common/block/include/intelblocks/cse.h
@@ -0,0 +1,50 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright (C) 2017 Intel Corp.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef SOC_INTEL_COMMON_CSE_H
+#define SOC_INTEL_COMMON_CSE_H
+
+#include <stdint.h>
+
+/* set up device for use in early boot enviroument with temp bar */
+void heci_init(uintptr_t bar);
+/*
+ * Receive message into buff not exceeding maxlen. Message is considered
+ * successfully received if a 'complete' indication is read from ME side
+ * and there was enough space in the buffer to fit that message. maxlen
+ * is updated with size of message that was received. Returns 0 on failure
+ * and 1 on success.
+ * In case of error heci_reset() may be requiered.
+ */
+int heci_receive(void *buff, size_t *maxlen);
+/*
+ * Send message msg of size len to host from host_addr to cse_addr.
+ * Returns 1 on success and 0 otherwise.
+ * In case of error heci_reset() may be requiered.
+ */
+int
+heci_send(const void *msg, size_t len, uint8_t host_addr, uint8_t cse_addr);
+/*
+ * Attempt device reset. This is useful and perhaps only thing left to do when
+ * CPU and CSE are out of sync or CSE fails to respond.
+ * Returns 0 on failure a 1 on success.
+ */
+int heci_reset(void);
+
+#define BIOS_HOST_ADDR							0x00
+#define HECI_MKHI_ADDR							0x07
+
+#endif // SOC_INTEL_COMMON_MSR_H