diff mbox series

[07/17] soc: qcom: ipa: the generic software interface

Message ID 20200228224204.17746-8-elder@linaro.org
State New
Headers show
Series [01/17] remoteproc: add IPA notification to q6v5 driver | expand

Commit Message

Alex Elder Feb. 28, 2020, 10:41 p.m. UTC
This patch includes "gsi.c", which implements the generic software
interface (GSI) for IPA.  The generic software interface abstracts
channels, which provide a means of transferring data either from the
AP to the IPA, or from the IPA to the AP.  A ring buffer of "transfer
elements" (TREs) is used to describe data transfers to perform.  The
AP writes a doorbell register associated with a channel to let it know
it has added new entries (for an AP->IPA channel) or has finished
processing entries (for an IPA->AP channel).

Each channel also has an event ring buffer, used by the IPA to
communicate information about events related to a channel (for
example, the completion of TREs).  The IPA writes its own doorbell
register, which triggers an interrupt on the AP, to signal that
new event information has arrived.

Signed-off-by: Alex Elder <elder@linaro.org>
---
 drivers/net/ipa/gsi.c | 2097 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 2097 insertions(+)
 create mode 100644 drivers/net/ipa/gsi.c
diff mbox series

Patch

diff --git a/drivers/net/ipa/gsi.c b/drivers/net/ipa/gsi.c
new file mode 100644
index 000000000000..f48d74f44592
--- /dev/null
+++ b/drivers/net/ipa/gsi.c
@@ -0,0 +1,2097 @@ 
+// SPDX-License-Identifier: GPL-2.0
+
+/* Copyright (c) 2015-2018, The Linux Foundation. All rights reserved.
+ * Copyright (C) 2018-2020 Linaro Ltd.
+ */
+
+#include <linux/types.h>
+#include <linux/bits.h>
+#include <linux/bitfield.h>
+#include <linux/mutex.h>
+#include <linux/completion.h>
+#include <linux/io.h>
+#include <linux/bug.h>
+#include <linux/interrupt.h>
+#include <linux/platform_device.h>
+#include <linux/netdevice.h>
+
+#include "gsi.h"
+#include "gsi_reg.h"
+#include "gsi_private.h"
+#include "gsi_trans.h"
+#include "ipa_gsi.h"
+#include "ipa_data.h"
+
+/**
+ * DOC: The IPA Generic Software Interface
+ *
+ * The generic software interface (GSI) is an integral component of the IPA,
+ * providing a well-defined communication layer between the AP subsystem
+ * and the IPA core.  The modem uses the GSI layer as well.
+ *
+ *	--------	     ---------
+ *	|      |	     |	     |
+ *	|  AP  +<---.	.----+ Modem |
+ *	|      +--. |	| .->+	     |
+ *	|      |  | |	| |  |	     |
+ *	--------  | |	| |  ---------
+ *		  v |	v |
+ *		--+-+---+-+--
+ *		|    GSI    |
+ *		|-----------|
+ *		|	    |
+ *		|    IPA    |
+ *		|	    |
+ *		-------------
+ *
+ * In the above diagram, the AP and Modem represent "execution environments"
+ * (EEs), which are independent operating environments that use the IPA for
+ * data transfer.
+ *
+ * Each EE uses a set of unidirectional GSI "channels," which allow transfer
+ * of data to or from the IPA.  A channel is implemented as a ring buffer,
+ * with a DRAM-resident array of "transfer elements" (TREs) available to
+ * describe transfers to or from other EEs through the IPA.  A transfer
+ * element can also contain an immediate command, requesting the IPA perform
+ * actions other than data transfer.
+ *
+ * Each TRE refers to a block of data--also located DRAM.  After writing one
+ * or more TREs to a channel, the writer (either the IPA or an EE) writes a
+ * doorbell register to inform the receiving side how many elements have
+ * been written.
+ *
+ * Each channel has a GSI "event ring" associated with it.  An event ring
+ * is implemented very much like a channel ring, but is always directed from
+ * the IPA to an EE.  The IPA notifies an EE (such as the AP) about channel
+ * events by adding an entry to the event ring associated with the channel.
+ * The GSI then writes its doorbell for the event ring, causing the target
+ * EE to be interrupted.  Each entry in an event ring contains a pointer
+ * to the channel TRE whose completion the event represents.
+ *
+ * Each TRE in a channel ring has a set of flags.  One flag indicates whether
+ * the completion of the transfer operation generates an entry (and possibly
+ * an interrupt) in the channel's event ring.  Other flags allow transfer
+ * elements to be chained together, forming a single logical transaction.
+ * TRE flags are used to control whether and when interrupts are generated
+ * to signal completion of channel transfers.
+ *
+ * Elements in channel and event rings are completed (or consumed) strictly
+ * in order.  Completion of one entry implies the completion of all preceding
+ * entries.  A single completion interrupt can therefore communicate the
+ * completion of many transfers.
+ *
+ * Note that all GSI registers are little-endian, which is the assumed
+ * endianness of I/O space accesses.  The accessor functions perform byte
+ * swapping if needed (i.e., for a big endian CPU).
+ */
+
+/* Delay period for interrupt moderation (in 32KHz IPA internal timer ticks) */
+#define GSI_EVT_RING_INT_MODT	(32 * 1) /* 1ms under 32KHz clock */
+
+#define GSI_CMD_TIMEOUT		5	/* seconds */
+
+#define GSI_MHI_EVENT_ID_START	10	/* 1st event id reserved for MHI */
+#define GSI_MHI_EVENT_ID_END	16	/* Last event id reserved for MHI */
+
+#define GSI_ISR_MAX_ITER	50	/* Detect interrupt storms */
+
+/* An entry in an event ring */
+struct gsi_event {
+	__le64 xfer_ptr;
+	__le16 len;
+	u8 reserved1;
+	u8 code;
+	__le16 reserved2;
+	u8 type;
+	u8 chid;
+};
+
+/* Hardware values from the error log register error code field */
+enum gsi_err_code {
+	GSI_INVALID_TRE_ERR			= 0x1,
+	GSI_OUT_OF_BUFFERS_ERR			= 0x2,
+	GSI_OUT_OF_RESOURCES_ERR		= 0x3,
+	GSI_UNSUPPORTED_INTER_EE_OP_ERR		= 0x4,
+	GSI_EVT_RING_EMPTY_ERR			= 0x5,
+	GSI_NON_ALLOCATED_EVT_ACCESS_ERR	= 0x6,
+	GSI_HWO_1_ERR				= 0x8,
+};
+
+/* Hardware values from the error log register error type field */
+enum gsi_err_type {
+	GSI_ERR_TYPE_GLOB	= 0x1,
+	GSI_ERR_TYPE_CHAN	= 0x2,
+	GSI_ERR_TYPE_EVT	= 0x3,
+};
+
+/* Hardware values used when programming an event ring */
+enum gsi_evt_chtype {
+	GSI_EVT_CHTYPE_MHI_EV	= 0x0,
+	GSI_EVT_CHTYPE_XHCI_EV	= 0x1,
+	GSI_EVT_CHTYPE_GPI_EV	= 0x2,
+	GSI_EVT_CHTYPE_XDCI_EV	= 0x3,
+};
+
+/* Hardware values used when programming a channel */
+enum gsi_channel_protocol {
+	GSI_CHANNEL_PROTOCOL_MHI	= 0x0,
+	GSI_CHANNEL_PROTOCOL_XHCI	= 0x1,
+	GSI_CHANNEL_PROTOCOL_GPI	= 0x2,
+	GSI_CHANNEL_PROTOCOL_XDCI	= 0x3,
+};
+
+/* Hardware values representing an event ring immediate command opcode */
+enum gsi_evt_cmd_opcode {
+	GSI_EVT_ALLOCATE	= 0x0,
+	GSI_EVT_RESET		= 0x9,
+	GSI_EVT_DE_ALLOC	= 0xa,
+};
+
+/* Hardware values representing a generic immediate command opcode */
+enum gsi_generic_cmd_opcode {
+	GSI_GENERIC_HALT_CHANNEL	= 0x1,
+	GSI_GENERIC_ALLOCATE_CHANNEL	= 0x2,
+};
+
+/* Hardware values representing a channel immediate command opcode */
+enum gsi_ch_cmd_opcode {
+	GSI_CH_ALLOCATE	= 0x0,
+	GSI_CH_START	= 0x1,
+	GSI_CH_STOP	= 0x2,
+	GSI_CH_RESET	= 0x9,
+	GSI_CH_DE_ALLOC	= 0xa,
+};
+
+/** gsi_channel_scratch_gpi - GPI protocol scratch register
+ * @max_outstanding_tre:
+ *	Defines the maximum number of TREs allowed in a single transaction
+ *	on a channel (in bytes).  This determines the amount of prefetch
+ *	performed by the hardware.  We configure this to equal the size of
+ *	the TLV FIFO for the channel.
+ * @outstanding_threshold:
+ *	Defines the threshold (in bytes) determining when the sequencer
+ *	should update the channel doorbell.  We configure this to equal
+ *	the size of two TREs.
+ */
+struct gsi_channel_scratch_gpi {
+	u64 reserved1;
+	u16 reserved2;
+	u16 max_outstanding_tre;
+	u16 reserved3;
+	u16 outstanding_threshold;
+};
+
+/** gsi_channel_scratch - channel scratch configuration area
+ *
+ * The exact interpretation of this register is protocol-specific.
+ * We only use GPI channels; see struct gsi_channel_scratch_gpi, above.
+ */
+union gsi_channel_scratch {
+	struct gsi_channel_scratch_gpi gpi;
+	struct {
+		u32 word1;
+		u32 word2;
+		u32 word3;
+		u32 word4;
+	} data;
+};
+
+/* Check things that can be validated at build time. */
+static void gsi_validate_build(void)
+{
+	/* This is used as a divisor */
+	BUILD_BUG_ON(!GSI_RING_ELEMENT_SIZE);
+
+	/* Code assumes the size of channel and event ring element are
+	 * the same (and fixed).  Make sure the size of an event ring
+	 * element is what's expected.
+	 */
+	BUILD_BUG_ON(sizeof(struct gsi_event) != GSI_RING_ELEMENT_SIZE);
+
+	/* Hardware requires a 2^n ring size.  We ensure the number of
+	 * elements in an event ring is a power of 2 elsewhere; this
+	 * ensure the elements themselves meet the requirement.
+	 */
+	BUILD_BUG_ON(!is_power_of_2(GSI_RING_ELEMENT_SIZE));
+
+	/* The channel element size must fit in this field */
+	BUILD_BUG_ON(GSI_RING_ELEMENT_SIZE > field_max(ELEMENT_SIZE_FMASK));
+
+	/* The event ring element size must fit in this field */
+	BUILD_BUG_ON(GSI_RING_ELEMENT_SIZE > field_max(EV_ELEMENT_SIZE_FMASK));
+}
+
+/* Return the channel id associated with a given channel */
+static u32 gsi_channel_id(struct gsi_channel *channel)
+{
+	return channel - &channel->gsi->channel[0];
+}
+
+static void gsi_irq_ieob_enable(struct gsi *gsi, u32 evt_ring_id)
+{
+	u32 val;
+
+	gsi->event_enable_bitmap |= BIT(evt_ring_id);
+	val = gsi->event_enable_bitmap;
+	iowrite32(val, gsi->virt + GSI_CNTXT_SRC_IEOB_IRQ_MSK_OFFSET);
+}
+
+static void gsi_isr_ieob_clear(struct gsi *gsi, u32 mask)
+{
+	iowrite32(mask, gsi->virt + GSI_CNTXT_SRC_IEOB_IRQ_CLR_OFFSET);
+}
+
+static void gsi_irq_ieob_disable(struct gsi *gsi, u32 evt_ring_id)
+{
+	u32 val;
+
+	gsi->event_enable_bitmap &= ~BIT(evt_ring_id);
+	val = gsi->event_enable_bitmap;
+	iowrite32(val, gsi->virt + GSI_CNTXT_SRC_IEOB_IRQ_MSK_OFFSET);
+}
+
+/* Enable all GSI_interrupt types */
+static void gsi_irq_enable(struct gsi *gsi)
+{
+	u32 val;
+
+	/* We don't use inter-EE channel or event interrupts */
+	val = GSI_CNTXT_TYPE_IRQ_MSK_ALL;
+	val &= ~MSK_INTER_EE_CH_CTRL_FMASK;
+	val &= ~MSK_INTER_EE_EV_CTRL_FMASK;
+	iowrite32(val, gsi->virt + GSI_CNTXT_TYPE_IRQ_MSK_OFFSET);
+
+	val = GENMASK(gsi->channel_count - 1, 0);
+	iowrite32(val, gsi->virt + GSI_CNTXT_SRC_CH_IRQ_MSK_OFFSET);
+
+	val = GENMASK(gsi->evt_ring_count - 1, 0);
+	iowrite32(val, gsi->virt + GSI_CNTXT_SRC_EV_CH_IRQ_MSK_OFFSET);
+
+	/* Each IEOB interrupt is enabled (later) as needed by channels */
+	iowrite32(0, gsi->virt + GSI_CNTXT_SRC_IEOB_IRQ_MSK_OFFSET);
+
+	val = GSI_CNTXT_GLOB_IRQ_ALL;
+	iowrite32(val, gsi->virt + GSI_CNTXT_GLOB_IRQ_EN_OFFSET);
+
+	/* Never enable GSI_BREAK_POINT */
+	val = GSI_CNTXT_GSI_IRQ_ALL & ~EN_BREAK_POINT_FMASK;
+	iowrite32(val, gsi->virt + GSI_CNTXT_GSI_IRQ_EN_OFFSET);
+}
+
+/* Disable all GSI_interrupt types */
+static void gsi_irq_disable(struct gsi *gsi)
+{
+	iowrite32(0, gsi->virt + GSI_CNTXT_GSI_IRQ_EN_OFFSET);
+	iowrite32(0, gsi->virt + GSI_CNTXT_GLOB_IRQ_EN_OFFSET);
+	iowrite32(0, gsi->virt + GSI_CNTXT_SRC_IEOB_IRQ_MSK_OFFSET);
+	iowrite32(0, gsi->virt + GSI_CNTXT_SRC_EV_CH_IRQ_MSK_OFFSET);
+	iowrite32(0, gsi->virt + GSI_CNTXT_SRC_CH_IRQ_MSK_OFFSET);
+	iowrite32(0, gsi->virt + GSI_CNTXT_TYPE_IRQ_MSK_OFFSET);
+}
+
+/* Return the virtual address associated with a ring index */
+void *gsi_ring_virt(struct gsi_ring *ring, u32 index)
+{
+	/* Note: index *must* be used modulo the ring count here */
+	return ring->virt + (index % ring->count) * GSI_RING_ELEMENT_SIZE;
+}
+
+/* Return the 32-bit DMA address associated with a ring index */
+static u32 gsi_ring_addr(struct gsi_ring *ring, u32 index)
+{
+	return (ring->addr & GENMASK(31, 0)) + index * GSI_RING_ELEMENT_SIZE;
+}
+
+/* Return the ring index of a 32-bit ring offset */
+static u32 gsi_ring_index(struct gsi_ring *ring, u32 offset)
+{
+	return (offset - gsi_ring_addr(ring, 0)) / GSI_RING_ELEMENT_SIZE;
+}
+
+/* Issue a GSI command by writing a value to a register, then wait for
+ * completion to be signaled.  Reports an error if the command times out.
+ * (Timeout is not expected, and suggests broken hardware.)
+ */
+static int
+gsi_command(struct gsi *gsi, u32 reg, u32 val, struct completion *completion)
+{
+	reinit_completion(completion);
+
+	iowrite32(val, gsi->virt + reg);
+	if (!wait_for_completion_timeout(completion, GSI_CMD_TIMEOUT * HZ))
+		return -ETIMEDOUT;
+
+	return 0;
+}
+
+/* Return the hardware's notion of the current state of an event ring */
+static enum gsi_evt_ring_state
+gsi_evt_ring_state(struct gsi *gsi, u32 evt_ring_id)
+{
+	u32 val;
+
+	val = ioread32(gsi->virt + GSI_EV_CH_E_CNTXT_0_OFFSET(evt_ring_id));
+
+	return u32_get_bits(val, EV_CHSTATE_FMASK);
+}
+
+/* Return whether an event ring's state is valid for an operation */
+static bool gsi_evt_ring_state_valid(struct gsi *gsi, u32 evt_ring_id,
+				     enum gsi_evt_cmd_opcode opcode)
+{
+	struct gsi_evt_ring *evt_ring = &gsi->evt_ring[evt_ring_id];
+	struct device *dev = evt_ring->channel->gsi->dev;
+	enum gsi_evt_ring_state state = evt_ring->state;
+	bool valid;
+
+	switch (opcode) {
+	case GSI_EVT_ALLOCATE:
+		valid = state == GSI_EVT_RING_STATE_NOT_ALLOCATED;
+		break;
+
+	case GSI_EVT_RESET:
+		valid = state == GSI_EVT_RING_STATE_ALLOCATED ||
+			state == GSI_EVT_RING_STATE_ERROR;
+		break;
+
+	case GSI_EVT_DE_ALLOC:
+		valid = state == GSI_EVT_RING_STATE_ALLOCATED;
+		break;
+
+	default:
+		dev_err(dev,
+			"event ring %u unrecognized state %u for opcode %u\n",
+			evt_ring_id, state, opcode);
+		return false;
+	}
+
+	if (!valid)
+		dev_err(dev,
+			"event ring %u unexpected state %u for opcode %u\n",
+			evt_ring_id, state, opcode);
+
+	return valid;
+}
+
+/* Issue an event ring command and wait for it to complete */
+static int evt_ring_command(struct gsi *gsi, u32 evt_ring_id,
+			    enum gsi_evt_cmd_opcode opcode)
+{
+	struct gsi_evt_ring *evt_ring = &gsi->evt_ring[evt_ring_id];
+	struct completion *completion = &evt_ring->completion;
+	u32 val;
+	int ret;
+
+	if (!gsi_evt_ring_state_valid(gsi, evt_ring_id, opcode))
+		return -EINVAL;
+
+	val = u32_encode_bits(evt_ring_id, EV_CHID_FMASK);
+	val |= u32_encode_bits(opcode, EV_OPCODE_FMASK);
+
+	ret = gsi_command(gsi, GSI_EV_CH_CMD_OFFSET, val, completion);
+	if (ret)
+		dev_err(gsi->dev,
+			"error %d issuing command %u to event ring %u\n",
+			ret, opcode, evt_ring_id);
+
+	return ret;
+}
+
+/* Allocate an event ring in NOT_ALLOCATED state */
+static int gsi_evt_ring_alloc_command(struct gsi *gsi, u32 evt_ring_id)
+{
+	struct gsi_evt_ring *evt_ring = &gsi->evt_ring[evt_ring_id];
+	int ret;
+
+	/* Get initial event ring state */
+	evt_ring->state = gsi_evt_ring_state(gsi, evt_ring_id);
+
+	ret = evt_ring_command(gsi, evt_ring_id, GSI_EVT_ALLOCATE);
+	if (!ret && evt_ring->state != GSI_EVT_RING_STATE_ALLOCATED) {
+		dev_err(gsi->dev, "bad event ring state (%u) after alloc\n",
+			evt_ring->state);
+		ret = -EIO;
+	}
+
+	return ret;
+}
+
+/* Reset a GSI event ring in ALLOCATED or ERROR state. */
+static void gsi_evt_ring_reset_command(struct gsi *gsi, u32 evt_ring_id)
+{
+	struct gsi_evt_ring *evt_ring = &gsi->evt_ring[evt_ring_id];
+	int ret;
+
+	ret = evt_ring_command(gsi, evt_ring_id, GSI_EVT_RESET);
+	if (!ret && evt_ring->state != GSI_EVT_RING_STATE_ALLOCATED)
+		dev_err(gsi->dev, "bad event ring state (%u) after reset\n",
+			evt_ring->state);
+}
+
+/* Issue a hardware de-allocation request for an allocated event ring */
+static void gsi_evt_ring_de_alloc_command(struct gsi *gsi, u32 evt_ring_id)
+{
+	struct gsi_evt_ring *evt_ring = &gsi->evt_ring[evt_ring_id];
+	int ret;
+
+	ret = evt_ring_command(gsi, evt_ring_id, GSI_EVT_DE_ALLOC);
+	if (!ret && evt_ring->state != GSI_EVT_RING_STATE_NOT_ALLOCATED)
+		dev_err(gsi->dev, "bad event ring state (%u) after dealloc\n",
+			evt_ring->state);
+}
+
+/* Return the hardware's notion of the current state of a channel */
+static enum gsi_channel_state
+gsi_channel_state(struct gsi *gsi, u32 channel_id)
+{
+	u32 val;
+
+	val = ioread32(gsi->virt + GSI_CH_C_CNTXT_0_OFFSET(channel_id));
+
+	return u32_get_bits(val, CHSTATE_FMASK);
+}
+
+/* Return whether a channel's state is valid for an operation */
+static bool gsi_channel_state_valid(struct gsi *gsi, u32 channel_id,
+				    enum gsi_ch_cmd_opcode opcode)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+	enum gsi_channel_state state = channel->state;
+	struct device *dev = channel->gsi->dev;
+	bool valid;
+
+	switch (opcode) {
+	case GSI_CH_ALLOCATE:
+		valid = state == GSI_CHANNEL_STATE_NOT_ALLOCATED;
+		break;
+
+	case GSI_CH_START:
+		valid = state == GSI_CHANNEL_STATE_ALLOCATED ||
+			state == GSI_CHANNEL_STATE_STOP_IN_PROC ||
+			state == GSI_CHANNEL_STATE_STOPPED;
+		break;
+
+	case GSI_CH_STOP:
+		valid = state == GSI_CHANNEL_STATE_STARTED ||
+			state == GSI_CHANNEL_STATE_STOP_IN_PROC ||
+			state == GSI_CHANNEL_STATE_ERROR;
+		break;
+
+	case GSI_CH_RESET:
+		valid = state == GSI_CHANNEL_STATE_STOPPED;
+		break;
+
+	case GSI_CH_DE_ALLOC:
+		valid = state == GSI_CHANNEL_STATE_ALLOCATED;
+		break;
+
+	default:
+		dev_err(dev,
+			"channel %u unrecognized state %u for opcode %u\n",
+			channel_id, state, opcode);
+		return false;
+	}
+
+	if (!valid)
+		dev_err(dev, "channel %u unexpected state %u for opcode %u\n",
+			channel_id, state, opcode);
+
+	return valid;
+}
+
+/* Issue a channel command and wait for it to complete */
+static int
+gsi_channel_command(struct gsi_channel *channel, enum gsi_ch_cmd_opcode opcode)
+{
+	struct completion *completion = &channel->completion;
+	u32 channel_id = gsi_channel_id(channel);
+	u32 val;
+	int ret;
+
+	if (!gsi_channel_state_valid(channel->gsi, channel_id, opcode))
+		return -EINVAL;
+
+	val = u32_encode_bits(channel_id, CH_CHID_FMASK);
+	val |= u32_encode_bits(opcode, CH_OPCODE_FMASK);
+
+	ret = gsi_command(channel->gsi, GSI_CH_CMD_OFFSET, val, completion);
+	if (ret)
+		dev_err(channel->gsi->dev,
+			"error %d issuing command %u to channel %u\n",
+			ret, opcode, channel_id);
+
+	return ret;
+}
+
+/* Allocate GSI channel in NOT_ALLOCATED state */
+static int gsi_channel_alloc_command(struct gsi *gsi, u32 channel_id)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+	int ret;
+
+	/* Get initial channel state */
+	channel->state = gsi_channel_state(gsi, channel_id);
+
+	ret = gsi_channel_command(channel, GSI_CH_ALLOCATE);
+	if (!ret && channel->state != GSI_CHANNEL_STATE_ALLOCATED) {
+		dev_err(gsi->dev, "bad channel state (%u) after alloc\n",
+			channel->state);
+		ret = -EIO;
+	}
+
+	return ret;
+}
+
+/* Start an ALLOCATED channel */
+static int gsi_channel_start_command(struct gsi_channel *channel)
+{
+	int ret;
+
+	ret = gsi_channel_command(channel, GSI_CH_START);
+	if (!ret && channel->state != GSI_CHANNEL_STATE_STARTED) {
+		dev_err(channel->gsi->dev,
+			"bad channel state (%u) after start\n",
+			channel->state);
+		ret = -EIO;
+	}
+
+	return ret;
+}
+
+/* Stop a GSI channel in STARTED state */
+static int gsi_channel_stop_command(struct gsi_channel *channel)
+{
+	int ret;
+
+	ret = gsi_channel_command(channel, GSI_CH_STOP);
+	if (ret || channel->state == GSI_CHANNEL_STATE_STOPPED)
+		return ret;
+
+	/* We may have to try again if stop is in progress */
+	if (channel->state == GSI_CHANNEL_STATE_STOP_IN_PROC)
+		return -EAGAIN;
+
+	dev_err(channel->gsi->dev, "bad channel state (%u) after stop\n",
+		channel->state);
+
+	return -EIO;
+}
+
+/* Reset a GSI channel in ALLOCATED or ERROR state. */
+static void gsi_channel_reset_command(struct gsi_channel *channel)
+{
+	int ret;
+
+	msleep(1);	/* A short delay is required before a RESET command */
+
+	ret = gsi_channel_command(channel, GSI_CH_RESET);
+	if (!ret && channel->state != GSI_CHANNEL_STATE_ALLOCATED)
+		dev_err(channel->gsi->dev,
+			"bad channel state (%u) after reset\n",
+			channel->state);
+}
+
+/* Deallocate an ALLOCATED GSI channel */
+static void gsi_channel_de_alloc_command(struct gsi *gsi, u32 channel_id)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+	int ret;
+
+	ret = gsi_channel_command(channel, GSI_CH_DE_ALLOC);
+	if (!ret && channel->state != GSI_CHANNEL_STATE_NOT_ALLOCATED)
+		dev_err(gsi->dev, "bad channel state (%u) after dealloc\n",
+			channel->state);
+}
+
+/* Ring an event ring doorbell, reporting the last entry processed by the AP.
+ * The index argument (modulo the ring count) is the first unfilled entry, so
+ * we supply one less than that with the doorbell.  Update the event ring
+ * index field with the value provided.
+ */
+static void gsi_evt_ring_doorbell(struct gsi *gsi, u32 evt_ring_id, u32 index)
+{
+	struct gsi_ring *ring = &gsi->evt_ring[evt_ring_id].ring;
+	u32 val;
+
+	ring->index = index;	/* Next unused entry */
+
+	/* Note: index *must* be used modulo the ring count here */
+	val = gsi_ring_addr(ring, (index - 1) % ring->count);
+	iowrite32(val, gsi->virt + GSI_EV_CH_E_DOORBELL_0_OFFSET(evt_ring_id));
+}
+
+/* Program an event ring for use */
+static void gsi_evt_ring_program(struct gsi *gsi, u32 evt_ring_id)
+{
+	struct gsi_evt_ring *evt_ring = &gsi->evt_ring[evt_ring_id];
+	size_t size = evt_ring->ring.count * GSI_RING_ELEMENT_SIZE;
+	u32 val;
+
+	val = u32_encode_bits(GSI_EVT_CHTYPE_GPI_EV, EV_CHTYPE_FMASK);
+	val |= EV_INTYPE_FMASK;
+	val |= u32_encode_bits(GSI_RING_ELEMENT_SIZE, EV_ELEMENT_SIZE_FMASK);
+	iowrite32(val, gsi->virt + GSI_EV_CH_E_CNTXT_0_OFFSET(evt_ring_id));
+
+	val = u32_encode_bits(size, EV_R_LENGTH_FMASK);
+	iowrite32(val, gsi->virt + GSI_EV_CH_E_CNTXT_1_OFFSET(evt_ring_id));
+
+	/* The context 2 and 3 registers store the low-order and
+	 * high-order 32 bits of the address of the event ring,
+	 * respectively.
+	 */
+	val = evt_ring->ring.addr & GENMASK(31, 0);
+	iowrite32(val, gsi->virt + GSI_EV_CH_E_CNTXT_2_OFFSET(evt_ring_id));
+
+	val = evt_ring->ring.addr >> 32;
+	iowrite32(val, gsi->virt + GSI_EV_CH_E_CNTXT_3_OFFSET(evt_ring_id));
+
+	/* Enable interrupt moderation by setting the moderation delay */
+	val = u32_encode_bits(GSI_EVT_RING_INT_MODT, MODT_FMASK);
+	val |= u32_encode_bits(1, MODC_FMASK);	/* comes from channel */
+	iowrite32(val, gsi->virt + GSI_EV_CH_E_CNTXT_8_OFFSET(evt_ring_id));
+
+	/* No MSI write data, and MSI address high and low address is 0 */
+	iowrite32(0, gsi->virt + GSI_EV_CH_E_CNTXT_9_OFFSET(evt_ring_id));
+	iowrite32(0, gsi->virt + GSI_EV_CH_E_CNTXT_10_OFFSET(evt_ring_id));
+	iowrite32(0, gsi->virt + GSI_EV_CH_E_CNTXT_11_OFFSET(evt_ring_id));
+
+	/* We don't need to get event read pointer updates */
+	iowrite32(0, gsi->virt + GSI_EV_CH_E_CNTXT_12_OFFSET(evt_ring_id));
+	iowrite32(0, gsi->virt + GSI_EV_CH_E_CNTXT_13_OFFSET(evt_ring_id));
+
+	/* Finally, tell the hardware we've completed event 0 (arbitrary) */
+	gsi_evt_ring_doorbell(gsi, evt_ring_id, 0);
+}
+
+/* Return the last (most recent) transaction completed on a channel. */
+static struct gsi_trans *gsi_channel_trans_last(struct gsi_channel *channel)
+{
+	struct gsi_trans_info *trans_info = &channel->trans_info;
+	struct gsi_trans *trans;
+
+	spin_lock_bh(&trans_info->spinlock);
+
+	if (!list_empty(&trans_info->complete))
+		trans = list_last_entry(&trans_info->complete,
+					struct gsi_trans, links);
+	else if (!list_empty(&trans_info->polled))
+		trans = list_last_entry(&trans_info->polled,
+					struct gsi_trans, links);
+	else
+		trans = NULL;
+
+	/* Caller will wait for this, so take a reference */
+	if (trans)
+		refcount_inc(&trans->refcount);
+
+	spin_unlock_bh(&trans_info->spinlock);
+
+	return trans;
+}
+
+/* Wait for transaction activity on a channel to complete */
+static void gsi_channel_trans_quiesce(struct gsi_channel *channel)
+{
+	struct gsi_trans *trans;
+
+	/* Get the last transaction, and wait for it to complete */
+	trans = gsi_channel_trans_last(channel);
+	if (trans) {
+		wait_for_completion(&trans->completion);
+		gsi_trans_free(trans);
+	}
+}
+
+/* Stop channel activity.  Transactions may not be allocated until thawed. */
+static void gsi_channel_freeze(struct gsi_channel *channel)
+{
+	gsi_channel_trans_quiesce(channel);
+
+	napi_disable(&channel->napi);
+
+	gsi_irq_ieob_disable(channel->gsi, channel->evt_ring_id);
+}
+
+/* Allow transactions to be used on the channel again. */
+static void gsi_channel_thaw(struct gsi_channel *channel)
+{
+	gsi_irq_ieob_enable(channel->gsi, channel->evt_ring_id);
+
+	napi_enable(&channel->napi);
+}
+
+/* Program a channel for use */
+static void gsi_channel_program(struct gsi_channel *channel, bool doorbell)
+{
+	size_t size = channel->tre_ring.count * GSI_RING_ELEMENT_SIZE;
+	u32 channel_id = gsi_channel_id(channel);
+	union gsi_channel_scratch scr = { };
+	struct gsi_channel_scratch_gpi *gpi;
+	struct gsi *gsi = channel->gsi;
+	u32 wrr_weight = 0;
+	u32 val;
+
+	/* Arbitrarily pick TRE 0 as the first channel element to use */
+	channel->tre_ring.index = 0;
+
+	/* We program all channels to use GPI protocol */
+	val = u32_encode_bits(GSI_CHANNEL_PROTOCOL_GPI, CHTYPE_PROTOCOL_FMASK);
+	if (channel->toward_ipa)
+		val |= CHTYPE_DIR_FMASK;
+	val |= u32_encode_bits(channel->evt_ring_id, ERINDEX_FMASK);
+	val |= u32_encode_bits(GSI_RING_ELEMENT_SIZE, ELEMENT_SIZE_FMASK);
+	iowrite32(val, gsi->virt + GSI_CH_C_CNTXT_0_OFFSET(channel_id));
+
+	val = u32_encode_bits(size, R_LENGTH_FMASK);
+	iowrite32(val, gsi->virt + GSI_CH_C_CNTXT_1_OFFSET(channel_id));
+
+	/* The context 2 and 3 registers store the low-order and
+	 * high-order 32 bits of the address of the channel ring,
+	 * respectively.
+	 */
+	val = channel->tre_ring.addr & GENMASK(31, 0);
+	iowrite32(val, gsi->virt + GSI_CH_C_CNTXT_2_OFFSET(channel_id));
+
+	val = channel->tre_ring.addr >> 32;
+	iowrite32(val, gsi->virt + GSI_CH_C_CNTXT_3_OFFSET(channel_id));
+
+	/* Command channel gets low weighted round-robin priority */
+	if (channel->command)
+		wrr_weight = field_max(WRR_WEIGHT_FMASK);
+	val = u32_encode_bits(wrr_weight, WRR_WEIGHT_FMASK);
+
+	/* Max prefetch is 1 segment (do not set MAX_PREFETCH_FMASK) */
+
+	/* Enable the doorbell engine if requested */
+	if (doorbell)
+		val |= USE_DB_ENG_FMASK;
+
+	if (!channel->use_prefetch)
+		val |= USE_ESCAPE_BUF_ONLY_FMASK;
+
+	iowrite32(val, gsi->virt + GSI_CH_C_QOS_OFFSET(channel_id));
+
+	/* Now update the scratch registers for GPI protocol */
+	gpi = &scr.gpi;
+	gpi->max_outstanding_tre = gsi_channel_trans_tre_max(gsi, channel_id) *
+					GSI_RING_ELEMENT_SIZE;
+	gpi->outstanding_threshold = 2 * GSI_RING_ELEMENT_SIZE;
+
+	val = scr.data.word1;
+	iowrite32(val, gsi->virt + GSI_CH_C_SCRATCH_0_OFFSET(channel_id));
+
+	val = scr.data.word2;
+	iowrite32(val, gsi->virt + GSI_CH_C_SCRATCH_1_OFFSET(channel_id));
+
+	val = scr.data.word3;
+	iowrite32(val, gsi->virt + GSI_CH_C_SCRATCH_2_OFFSET(channel_id));
+
+	/* We must preserve the upper 16 bits of the last scratch register.
+	 * The next sequence assumes those bits remain unchanged between the
+	 * read and the write.
+	 */
+	val = ioread32(gsi->virt + GSI_CH_C_SCRATCH_3_OFFSET(channel_id));
+	val = (scr.data.word4 & GENMASK(31, 16)) | (val & GENMASK(15, 0));
+	iowrite32(val, gsi->virt + GSI_CH_C_SCRATCH_3_OFFSET(channel_id));
+
+	/* All done! */
+}
+
+static void gsi_channel_deprogram(struct gsi_channel *channel)
+{
+	/* Nothing to do */
+}
+
+/* Start an allocated GSI channel */
+int gsi_channel_start(struct gsi *gsi, u32 channel_id)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+	u32 evt_ring_id = channel->evt_ring_id;
+	int ret;
+
+	mutex_lock(&gsi->mutex);
+
+	ret =  gsi_channel_start_command(channel);
+
+	mutex_unlock(&gsi->mutex);
+
+	/* Clear the channel's event ring interrupt in case it's pending */
+	gsi_isr_ieob_clear(gsi, BIT(evt_ring_id));
+
+	gsi_channel_thaw(channel);
+
+	return 0;
+}
+
+/* Stop a started channel */
+int gsi_channel_stop(struct gsi *gsi, u32 channel_id)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+	int ret;
+
+	gsi_channel_freeze(channel);
+
+	/* Channel could have entered STOPPED state since last call if the
+	 * STOP command timed out.  We won't stop a channel if stopping it
+	 * was successful previously (so we still want the freeze above).
+	 */
+	if (channel->state == GSI_CHANNEL_STATE_STOPPED)
+		return 0;
+
+	mutex_lock(&gsi->mutex);
+
+	ret = gsi_channel_stop_command(channel);
+
+	mutex_unlock(&gsi->mutex);
+
+	/* Thaw the channel if we need to retry (or on error) */
+	if (ret)
+		gsi_channel_thaw(channel);
+
+	return ret;
+}
+
+/* Reset and reconfigure a channel (possibly leaving doorbell disabled) */
+void gsi_channel_reset(struct gsi *gsi, u32 channel_id, bool db_enable)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+
+	mutex_lock(&gsi->mutex);
+
+	/* Due to a hardware quirk we need to reset RX channels twice. */
+	gsi_channel_reset_command(channel);
+	if (!channel->toward_ipa)
+		gsi_channel_reset_command(channel);
+
+	gsi_channel_program(channel, db_enable);
+	gsi_channel_trans_cancel_pending(channel);
+
+	mutex_unlock(&gsi->mutex);
+}
+
+/* Stop a STARTED channel for suspend (only stop if RX and requested) */
+int gsi_channel_suspend(struct gsi *gsi, u32 channel_id, bool stop)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+
+	if (stop && !channel->toward_ipa)
+		return gsi_channel_stop(gsi, channel_id);
+
+	gsi_channel_freeze(channel);
+
+	return 0;
+}
+
+/* Resume a suspended channel (starting wll be requested if STOPPED) */
+int gsi_channel_resume(struct gsi *gsi, u32 channel_id, bool start)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+
+	if (start && !channel->toward_ipa)
+		return gsi_channel_start(gsi, channel_id);
+
+	gsi_channel_thaw(channel);
+
+	return 0;
+}
+
+/**
+ * gsi_channel_tx_queued() - Report queued TX transfers for a channel
+ * @channel:	Channel for which to report
+ *
+ * Report to the network stack the number of bytes and transactions that
+ * have been queued to hardware since last call.  This and the next function
+ * supply information used by the network stack for throttling.
+ *
+ * For each channel we track the number of transactions used and bytes of
+ * data those transactions represent.  We also track what those values are
+ * each time this function is called.  Subtracting the two tells us
+ * the number of bytes and transactions that have been added between
+ * successive calls.
+ *
+ * Calling this each time we ring the channel doorbell allows us to
+ * provide accurate information to the network stack about how much
+ * work we've given the hardware at any point in time.
+ */
+void gsi_channel_tx_queued(struct gsi_channel *channel)
+{
+	u32 trans_count;
+	u32 byte_count;
+
+	byte_count = channel->byte_count - channel->queued_byte_count;
+	trans_count = channel->trans_count - channel->queued_trans_count;
+	channel->queued_byte_count = channel->byte_count;
+	channel->queued_trans_count = channel->trans_count;
+
+	ipa_gsi_channel_tx_queued(channel->gsi, gsi_channel_id(channel),
+				  trans_count, byte_count);
+}
+
+/**
+ * gsi_channel_tx_update() - Report completed TX transfers
+ * @channel:	Channel that has completed transmitting packets
+ * @trans:	Last transation known to be complete
+ *
+ * Compute the number of transactions and bytes that have been transferred
+ * over a TX channel since the given transaction was committed.  Report this
+ * information to the network stack.
+ *
+ * At the time a transaction is committed, we record its channel's
+ * committed transaction and byte counts *in the transaction*.
+ * Completions are signaled by the hardware with an interrupt, and
+ * we can determine the latest completed transaction at that time.
+ *
+ * The difference between the byte/transaction count recorded in
+ * the transaction and the count last time we recorded a completion
+ * tells us exactly how much data has been transferred between
+ * completions.
+ *
+ * Calling this each time we learn of a newly-completed transaction
+ * allows us to provide accurate information to the network stack
+ * about how much work has been completed by the hardware at a given
+ * point in time.
+ */
+static void
+gsi_channel_tx_update(struct gsi_channel *channel, struct gsi_trans *trans)
+{
+	u64 byte_count = trans->byte_count + trans->len;
+	u64 trans_count = trans->trans_count + 1;
+
+	byte_count -= channel->compl_byte_count;
+	channel->compl_byte_count += byte_count;
+	trans_count -= channel->compl_trans_count;
+	channel->compl_trans_count += trans_count;
+
+	ipa_gsi_channel_tx_completed(channel->gsi, gsi_channel_id(channel),
+				     trans_count, byte_count);
+}
+
+/* Channel control interrupt handler */
+static void gsi_isr_chan_ctrl(struct gsi *gsi)
+{
+	u32 channel_mask;
+
+	channel_mask = ioread32(gsi->virt + GSI_CNTXT_SRC_CH_IRQ_OFFSET);
+	iowrite32(channel_mask, gsi->virt + GSI_CNTXT_SRC_CH_IRQ_CLR_OFFSET);
+
+	while (channel_mask) {
+		u32 channel_id = __ffs(channel_mask);
+		struct gsi_channel *channel;
+
+		channel_mask ^= BIT(channel_id);
+
+		channel = &gsi->channel[channel_id];
+		channel->state = gsi_channel_state(gsi, channel_id);
+
+		complete(&channel->completion);
+	}
+}
+
+/* Event ring control interrupt handler */
+static void gsi_isr_evt_ctrl(struct gsi *gsi)
+{
+	u32 event_mask;
+
+	event_mask = ioread32(gsi->virt + GSI_CNTXT_SRC_EV_CH_IRQ_OFFSET);
+	iowrite32(event_mask, gsi->virt + GSI_CNTXT_SRC_EV_CH_IRQ_CLR_OFFSET);
+
+	while (event_mask) {
+		u32 evt_ring_id = __ffs(event_mask);
+		struct gsi_evt_ring *evt_ring;
+
+		event_mask ^= BIT(evt_ring_id);
+
+		evt_ring = &gsi->evt_ring[evt_ring_id];
+		evt_ring->state = gsi_evt_ring_state(gsi, evt_ring_id);
+
+		complete(&evt_ring->completion);
+	}
+}
+
+/* Global channel error interrupt handler */
+static void
+gsi_isr_glob_chan_err(struct gsi *gsi, u32 err_ee, u32 channel_id, u32 code)
+{
+	if (code == GSI_OUT_OF_RESOURCES_ERR) {
+		dev_err(gsi->dev, "channel %u out of resources\n", channel_id);
+		complete(&gsi->channel[channel_id].completion);
+		return;
+	}
+
+	/* Report, but otherwise ignore all other error codes */
+	dev_err(gsi->dev, "channel %u global error ee 0x%08x code 0x%08x\n",
+		channel_id, err_ee, code);
+}
+
+/* Global event error interrupt handler */
+static void
+gsi_isr_glob_evt_err(struct gsi *gsi, u32 err_ee, u32 evt_ring_id, u32 code)
+{
+	if (code == GSI_OUT_OF_RESOURCES_ERR) {
+		struct gsi_evt_ring *evt_ring = &gsi->evt_ring[evt_ring_id];
+		u32 channel_id = gsi_channel_id(evt_ring->channel);
+
+		complete(&evt_ring->completion);
+		dev_err(gsi->dev, "evt_ring for channel %u out of resources\n",
+			channel_id);
+		return;
+	}
+
+	/* Report, but otherwise ignore all other error codes */
+	dev_err(gsi->dev, "event ring %u global error ee %u code 0x%08x\n",
+		evt_ring_id, err_ee, code);
+}
+
+/* Global error interrupt handler */
+static void gsi_isr_glob_err(struct gsi *gsi)
+{
+	enum gsi_err_type type;
+	enum gsi_err_code code;
+	u32 which;
+	u32 val;
+	u32 ee;
+
+	/* Get the logged error, then reinitialize the log */
+	val = ioread32(gsi->virt + GSI_ERROR_LOG_OFFSET);
+	iowrite32(0, gsi->virt + GSI_ERROR_LOG_OFFSET);
+	iowrite32(~0, gsi->virt + GSI_ERROR_LOG_CLR_OFFSET);
+
+	ee = u32_get_bits(val, ERR_EE_FMASK);
+	which = u32_get_bits(val, ERR_VIRT_IDX_FMASK);
+	type = u32_get_bits(val, ERR_TYPE_FMASK);
+	code = u32_get_bits(val, ERR_CODE_FMASK);
+
+	if (type == GSI_ERR_TYPE_CHAN)
+		gsi_isr_glob_chan_err(gsi, ee, which, code);
+	else if (type == GSI_ERR_TYPE_EVT)
+		gsi_isr_glob_evt_err(gsi, ee, which, code);
+	else	/* type GSI_ERR_TYPE_GLOB should be fatal */
+		dev_err(gsi->dev, "unexpected global error 0x%08x\n", type);
+}
+
+/* Generic EE interrupt handler */
+static void gsi_isr_gp_int1(struct gsi *gsi)
+{
+	u32 result;
+	u32 val;
+
+	val = ioread32(gsi->virt + GSI_CNTXT_SCRATCH_0_OFFSET);
+	result = u32_get_bits(val, GENERIC_EE_RESULT_FMASK);
+	if (result != GENERIC_EE_SUCCESS_FVAL)
+		dev_err(gsi->dev, "global INT1 generic result %u\n", result);
+
+	complete(&gsi->completion);
+}
+/* Inter-EE interrupt handler */
+static void gsi_isr_glob_ee(struct gsi *gsi)
+{
+	u32 val;
+
+	val = ioread32(gsi->virt + GSI_CNTXT_GLOB_IRQ_STTS_OFFSET);
+
+	if (val & ERROR_INT_FMASK)
+		gsi_isr_glob_err(gsi);
+
+	iowrite32(val, gsi->virt + GSI_CNTXT_GLOB_IRQ_CLR_OFFSET);
+
+	val &= ~ERROR_INT_FMASK;
+
+	if (val & EN_GP_INT1_FMASK) {
+		val ^= EN_GP_INT1_FMASK;
+		gsi_isr_gp_int1(gsi);
+	}
+
+	if (val)
+		dev_err(gsi->dev, "unexpected global interrupt 0x%08x\n", val);
+}
+
+/* I/O completion interrupt event */
+static void gsi_isr_ieob(struct gsi *gsi)
+{
+	u32 event_mask;
+
+	event_mask = ioread32(gsi->virt + GSI_CNTXT_SRC_IEOB_IRQ_OFFSET);
+	gsi_isr_ieob_clear(gsi, event_mask);
+
+	while (event_mask) {
+		u32 evt_ring_id = __ffs(event_mask);
+
+		event_mask ^= BIT(evt_ring_id);
+
+		gsi_irq_ieob_disable(gsi, evt_ring_id);
+		napi_schedule(&gsi->evt_ring[evt_ring_id].channel->napi);
+	}
+}
+
+/* General event interrupts represent serious problems, so report them */
+static void gsi_isr_general(struct gsi *gsi)
+{
+	struct device *dev = gsi->dev;
+	u32 val;
+
+	val = ioread32(gsi->virt + GSI_CNTXT_GSI_IRQ_STTS_OFFSET);
+	iowrite32(val, gsi->virt + GSI_CNTXT_GSI_IRQ_CLR_OFFSET);
+
+	if (val)
+		dev_err(dev, "unexpected general interrupt 0x%08x\n", val);
+}
+
+/**
+ * gsi_isr() - Top level GSI interrupt service routine
+ * @irq:	Interrupt number (ignored)
+ * @dev_id:	GSI pointer supplied to request_irq()
+ *
+ * This is the main handler function registered for the GSI IRQ. Each type
+ * of interrupt has a separate handler function that is called from here.
+ */
+static irqreturn_t gsi_isr(int irq, void *dev_id)
+{
+	struct gsi *gsi = dev_id;
+	u32 intr_mask;
+	u32 cnt = 0;
+
+	while ((intr_mask = ioread32(gsi->virt + GSI_CNTXT_TYPE_IRQ_OFFSET))) {
+		/* intr_mask contains bitmask of pending GSI interrupts */
+		do {
+			u32 gsi_intr = BIT(__ffs(intr_mask));
+
+			intr_mask ^= gsi_intr;
+
+			switch (gsi_intr) {
+			case CH_CTRL_FMASK:
+				gsi_isr_chan_ctrl(gsi);
+				break;
+			case EV_CTRL_FMASK:
+				gsi_isr_evt_ctrl(gsi);
+				break;
+			case GLOB_EE_FMASK:
+				gsi_isr_glob_ee(gsi);
+				break;
+			case IEOB_FMASK:
+				gsi_isr_ieob(gsi);
+				break;
+			case GENERAL_FMASK:
+				gsi_isr_general(gsi);
+				break;
+			default:
+				dev_err(gsi->dev,
+					"%s: unrecognized type 0x%08x\n",
+					__func__, gsi_intr);
+				break;
+			}
+		} while (intr_mask);
+
+		if (++cnt > GSI_ISR_MAX_ITER) {
+			dev_err(gsi->dev, "interrupt flood\n");
+			break;
+		}
+	}
+
+	return IRQ_HANDLED;
+}
+
+/* Return the transaction associated with a transfer completion event */
+static struct gsi_trans *gsi_event_trans(struct gsi_channel *channel,
+					 struct gsi_event *event)
+{
+	u32 tre_offset;
+	u32 tre_index;
+
+	/* Event xfer_ptr records the TRE it's associated with */
+	tre_offset = le64_to_cpu(event->xfer_ptr) & GENMASK(31, 0);
+	tre_index = gsi_ring_index(&channel->tre_ring, tre_offset);
+
+	return gsi_channel_trans_mapped(channel, tre_index);
+}
+
+/**
+ * gsi_evt_ring_rx_update() - Record lengths of received data
+ * @evt_ring:	Event ring associated with channel that received packets
+ * @index:	Event index in ring reported by hardware
+ *
+ * Events for RX channels contain the actual number of bytes received into
+ * the buffer.  Every event has a transaction associated with it, and here
+ * we update transactions to record their actual received lengths.
+ *
+ * This function is called whenever we learn that the GSI hardware has filled
+ * new events since the last time we checked.  The ring's index field tells
+ * the first entry in need of processing.  The index provided is the
+ * first *unfilled* event in the ring (following the last filled one).
+ *
+ * Events are sequential within the event ring, and transactions are
+ * sequential within the transaction pool.
+ *
+ * Note that @index always refers to an element *within* the event ring.
+ */
+static void gsi_evt_ring_rx_update(struct gsi_evt_ring *evt_ring, u32 index)
+{
+	struct gsi_channel *channel = evt_ring->channel;
+	struct gsi_ring *ring = &evt_ring->ring;
+	struct gsi_trans_info *trans_info;
+	struct gsi_event *event_done;
+	struct gsi_event *event;
+	struct gsi_trans *trans;
+	u32 byte_count = 0;
+	u32 old_index;
+	u32 event_avail;
+
+	trans_info = &channel->trans_info;
+
+	/* We'll start with the oldest un-processed event.  RX channels
+	 * replenish receive buffers in single-TRE transactions, so we
+	 * can just map that event to its transaction.  Transactions
+	 * associated with completion events are consecutive.
+	 */
+	old_index = ring->index;
+	event = gsi_ring_virt(ring, old_index);
+	trans = gsi_event_trans(channel, event);
+
+	/* Compute the number of events to process before we wrap,
+	 * and determine when we'll be done processing events.
+	 */
+	event_avail = ring->count - old_index % ring->count;
+	event_done = gsi_ring_virt(ring, index);
+	do {
+		trans->len = __le16_to_cpu(event->len);
+		byte_count += trans->len;
+
+		/* Move on to the next event and transaction */
+		if (--event_avail)
+			event++;
+		else
+			event = gsi_ring_virt(ring, 0);
+		trans = gsi_trans_pool_next(&trans_info->pool, trans);
+	} while (event != event_done);
+
+	/* We record RX bytes when they are received */
+	channel->byte_count += byte_count;
+	channel->trans_count++;
+}
+
+/* Initialize a ring, including allocating DMA memory for its entries */
+static int gsi_ring_alloc(struct gsi *gsi, struct gsi_ring *ring, u32 count)
+{
+	size_t size = count * GSI_RING_ELEMENT_SIZE;
+	struct device *dev = gsi->dev;
+	dma_addr_t addr;
+
+	/* Hardware requires a 2^n ring size, with alignment equal to size */
+	ring->virt = dma_alloc_coherent(dev, size, &addr, GFP_KERNEL);
+	if (ring->virt && addr % size) {
+		dma_free_coherent(dev, size, ring->virt, ring->addr);
+		dev_err(dev, "unable to alloc 0x%zx-aligned ring buffer\n",
+				size);
+		return -EINVAL;	/* Not a good error value, but distinct */
+	} else if (!ring->virt) {
+		return -ENOMEM;
+	}
+	ring->addr = addr;
+	ring->count = count;
+
+	return 0;
+}
+
+/* Free a previously-allocated ring */
+static void gsi_ring_free(struct gsi *gsi, struct gsi_ring *ring)
+{
+	size_t size = ring->count * GSI_RING_ELEMENT_SIZE;
+
+	dma_free_coherent(gsi->dev, size, ring->virt, ring->addr);
+}
+
+/* Allocate an available event ring id */
+static int gsi_evt_ring_id_alloc(struct gsi *gsi)
+{
+	u32 evt_ring_id;
+
+	if (gsi->event_bitmap == ~0U) {
+		dev_err(gsi->dev, "event rings exhausted\n");
+		return -ENOSPC;
+	}
+
+	evt_ring_id = ffz(gsi->event_bitmap);
+	gsi->event_bitmap |= BIT(evt_ring_id);
+
+	return (int)evt_ring_id;
+}
+
+/* Free a previously-allocated event ring id */
+static void gsi_evt_ring_id_free(struct gsi *gsi, u32 evt_ring_id)
+{
+	gsi->event_bitmap &= ~BIT(evt_ring_id);
+}
+
+/* Ring a channel doorbell, reporting the first un-filled entry */
+void gsi_channel_doorbell(struct gsi_channel *channel)
+{
+	struct gsi_ring *tre_ring = &channel->tre_ring;
+	u32 channel_id = gsi_channel_id(channel);
+	struct gsi *gsi = channel->gsi;
+	u32 val;
+
+	/* Note: index *must* be used modulo the ring count here */
+	val = gsi_ring_addr(tre_ring, tre_ring->index % tre_ring->count);
+	iowrite32(val, gsi->virt + GSI_CH_C_DOORBELL_0_OFFSET(channel_id));
+}
+
+/* Consult hardware, move any newly completed transactions to completed list */
+static void gsi_channel_update(struct gsi_channel *channel)
+{
+	u32 evt_ring_id = channel->evt_ring_id;
+	struct gsi *gsi = channel->gsi;
+	struct gsi_evt_ring *evt_ring;
+	struct gsi_trans *trans;
+	struct gsi_ring *ring;
+	u32 offset;
+	u32 index;
+
+	evt_ring = &gsi->evt_ring[evt_ring_id];
+	ring = &evt_ring->ring;
+
+	/* See if there's anything new to process; if not, we're done.  Note
+	 * that index always refers to an entry *within* the event ring.
+	 */
+	offset = GSI_EV_CH_E_CNTXT_4_OFFSET(evt_ring_id);
+	index = gsi_ring_index(ring, ioread32(gsi->virt + offset));
+	if (index == ring->index % ring->count)
+		return;
+
+	/* Get the transaction for the latest completed event.  Take a
+	 * reference to keep it from completing before we give the events
+	 * for this and previous transactions back to the hardware.
+	 */
+	trans = gsi_event_trans(channel, gsi_ring_virt(ring, index - 1));
+	refcount_inc(&trans->refcount);
+
+	/* For RX channels, update each completed transaction with the number
+	 * of bytes that were actually received.  For TX channels, report
+	 * the number of transactions and bytes this completion represents
+	 * up the network stack.
+	 */
+	if (channel->toward_ipa)
+		gsi_channel_tx_update(channel, trans);
+	else
+		gsi_evt_ring_rx_update(evt_ring, index);
+
+	gsi_trans_move_complete(trans);
+
+	/* Tell the hardware we've handled these events */
+	gsi_evt_ring_doorbell(channel->gsi, channel->evt_ring_id, index);
+
+	gsi_trans_free(trans);
+}
+
+/**
+ * gsi_channel_poll_one() - Return a single completed transaction on a channel
+ * @channel:	Channel to be polled
+ *
+ * @Return:	 Transaction pointer, or null if none are available
+ *
+ * This function returns the first entry on a channel's completed transaction
+ * list.  If that list is empty, the hardware is consulted to determine
+ * whether any new transactions have completed.  If so, they're moved to the
+ * completed list and the new first entry is returned.  If there are no more
+ * completed transactions, a null pointer is returned.
+ */
+static struct gsi_trans *gsi_channel_poll_one(struct gsi_channel *channel)
+{
+	struct gsi_trans *trans;
+
+	/* Get the first transaction from the completed list */
+	trans = gsi_channel_trans_complete(channel);
+	if (!trans) {
+		/* List is empty; see if there's more to do */
+		gsi_channel_update(channel);
+		trans = gsi_channel_trans_complete(channel);
+	}
+
+	if (trans)
+		gsi_trans_move_polled(trans);
+
+	return trans;
+}
+
+/**
+ * gsi_channel_poll() - NAPI poll function for a channel
+ * @napi:	NAPI structure for the channel
+ * @budget:	Budget supplied by NAPI core
+
+ * @Return:	 Number of items polled (<= budget)
+ *
+ * Single transactions completed by hardware are polled until either
+ * the budget is exhausted, or there are no more.  Each transaction
+ * polled is passed to gsi_trans_complete(), to perform remaining
+ * completion processing and retire/free the transaction.
+ */
+static int gsi_channel_poll(struct napi_struct *napi, int budget)
+{
+	struct gsi_channel *channel;
+	int count = 0;
+
+	channel = container_of(napi, struct gsi_channel, napi);
+	while (count < budget) {
+		struct gsi_trans *trans;
+
+		trans = gsi_channel_poll_one(channel);
+		if (!trans)
+			break;
+		gsi_trans_complete(trans);
+	}
+
+	if (count < budget) {
+		napi_complete(&channel->napi);
+		gsi_irq_ieob_enable(channel->gsi, channel->evt_ring_id);
+	}
+
+	return count;
+}
+
+/* The event bitmap represents which event ids are available for allocation.
+ * Set bits are not available, clear bits can be used.  This function
+ * initializes the map so all events supported by the hardware are available,
+ * then precludes any reserved events from being allocated.
+ */
+static u32 gsi_event_bitmap_init(u32 evt_ring_max)
+{
+	u32 event_bitmap = GENMASK(BITS_PER_LONG - 1, evt_ring_max);
+
+	event_bitmap |= GENMASK(GSI_MHI_EVENT_ID_END, GSI_MHI_EVENT_ID_START);
+
+	return event_bitmap;
+}
+
+/* Setup function for event rings */
+static void gsi_evt_ring_setup(struct gsi *gsi)
+{
+	/* Nothing to do */
+}
+
+/* Inverse of gsi_evt_ring_setup() */
+static void gsi_evt_ring_teardown(struct gsi *gsi)
+{
+	/* Nothing to do */
+}
+
+/* Setup function for a single channel */
+static int gsi_channel_setup_one(struct gsi *gsi, u32 channel_id,
+				 bool db_enable)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+	u32 evt_ring_id = channel->evt_ring_id;
+	int ret;
+
+	if (!channel->gsi)
+		return 0;	/* Ignore uninitialized channels */
+
+	ret = gsi_evt_ring_alloc_command(gsi, evt_ring_id);
+	if (ret)
+		return ret;
+
+	gsi_evt_ring_program(gsi, evt_ring_id);
+
+	ret = gsi_channel_alloc_command(gsi, channel_id);
+	if (ret)
+		goto err_evt_ring_de_alloc;
+
+	gsi_channel_program(channel, db_enable);
+
+	if (channel->toward_ipa)
+		netif_tx_napi_add(&gsi->dummy_dev, &channel->napi,
+				  gsi_channel_poll, NAPI_POLL_WEIGHT);
+	else
+		netif_napi_add(&gsi->dummy_dev, &channel->napi,
+			       gsi_channel_poll, NAPI_POLL_WEIGHT);
+
+	return 0;
+
+err_evt_ring_de_alloc:
+	/* We've done nothing with the event ring yet so don't reset */
+	gsi_evt_ring_de_alloc_command(gsi, evt_ring_id);
+
+	return ret;
+}
+
+/* Inverse of gsi_channel_setup_one() */
+static void gsi_channel_teardown_one(struct gsi *gsi, u32 channel_id)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+	u32 evt_ring_id = channel->evt_ring_id;
+
+	if (!channel->gsi)
+		return;		/* Ignore uninitialized channels */
+
+	netif_napi_del(&channel->napi);
+
+	gsi_channel_deprogram(channel);
+	gsi_channel_de_alloc_command(gsi, channel_id);
+	gsi_evt_ring_reset_command(gsi, evt_ring_id);
+	gsi_evt_ring_de_alloc_command(gsi, evt_ring_id);
+}
+
+static int gsi_generic_command(struct gsi *gsi, u32 channel_id,
+			       enum gsi_generic_cmd_opcode opcode)
+{
+	struct completion *completion = &gsi->completion;
+	u32 val;
+	u32 ret;
+
+	val = u32_encode_bits(opcode, GENERIC_OPCODE_FMASK);
+	val |= u32_encode_bits(channel_id, GENERIC_CHID_FMASK);
+	val |= u32_encode_bits(GSI_EE_MODEM, GENERIC_EE_FMASK);
+
+	ret = gsi_command(gsi, GSI_GENERIC_CMD_OFFSET, val, completion);
+	if (ret)
+		dev_err(gsi->dev,
+			"error %d issuing generic command %u for channel %u\n",
+			ret, opcode, channel_id);
+
+	return ret;
+}
+
+static int gsi_modem_channel_alloc(struct gsi *gsi, u32 channel_id)
+{
+	return gsi_generic_command(gsi, channel_id,
+				   GSI_GENERIC_ALLOCATE_CHANNEL);
+}
+
+static void gsi_modem_channel_halt(struct gsi *gsi, u32 channel_id)
+{
+	int ret;
+
+	ret = gsi_generic_command(gsi, channel_id, GSI_GENERIC_HALT_CHANNEL);
+	if (ret)
+		dev_err(gsi->dev, "error %d halting modem channel %u\n",
+			channel_id);
+}
+
+/* Setup function for channels */
+static int gsi_channel_setup(struct gsi *gsi, bool db_enable)
+{
+	u32 channel_id = 0;
+	u32 mask;
+	int ret;
+
+	gsi_evt_ring_setup(gsi);
+	gsi_irq_enable(gsi);
+
+	mutex_lock(&gsi->mutex);
+
+	do {
+		ret = gsi_channel_setup_one(gsi, channel_id, db_enable);
+		if (ret)
+			goto err_unwind;
+	} while (++channel_id < gsi->channel_count);
+
+	/* Make sure no channels were defined that hardware does not support */
+	while (channel_id < GSI_CHANNEL_COUNT_MAX) {
+		struct gsi_channel *channel = &gsi->channel[channel_id++];
+
+		if (!channel->gsi)
+			continue;	/* Ignore uninitialized channels */
+
+		dev_err(gsi->dev, "channel %u not supported by hardware\n",
+			channel_id - 1);
+		channel_id = gsi->channel_count;
+		goto err_unwind;
+	}
+
+	/* Allocate modem channels if necessary */
+	mask = gsi->modem_channel_bitmap;
+	while (mask) {
+		u32 modem_channel_id = __ffs(mask);
+
+		ret = gsi_modem_channel_alloc(gsi, modem_channel_id);
+		if (ret)
+			goto err_unwind_modem;
+
+		/* Clear bit from mask only after success (for unwind) */
+		mask ^= BIT(modem_channel_id);
+	}
+
+	mutex_unlock(&gsi->mutex);
+
+	return 0;
+
+err_unwind_modem:
+	/* Compute which modem channels need to be deallocated */
+	mask ^= gsi->modem_channel_bitmap;
+	while (mask) {
+		u32 channel_id = __fls(mask);
+
+		mask ^= BIT(channel_id);
+
+		gsi_modem_channel_halt(gsi, channel_id);
+	}
+
+err_unwind:
+	while (channel_id--)
+		gsi_channel_teardown_one(gsi, channel_id);
+
+	mutex_unlock(&gsi->mutex);
+
+	gsi_irq_disable(gsi);
+	gsi_evt_ring_teardown(gsi);
+
+	return ret;
+}
+
+/* Inverse of gsi_channel_setup() */
+static void gsi_channel_teardown(struct gsi *gsi)
+{
+	u32 mask = gsi->modem_channel_bitmap;
+	u32 channel_id;
+
+	mutex_lock(&gsi->mutex);
+
+	while (mask) {
+		u32 channel_id = __fls(mask);
+
+		mask ^= BIT(channel_id);
+
+		gsi_modem_channel_halt(gsi, channel_id);
+	}
+
+	channel_id = gsi->channel_count - 1;
+	do
+		gsi_channel_teardown_one(gsi, channel_id);
+	while (channel_id--);
+
+	mutex_unlock(&gsi->mutex);
+
+	gsi_irq_disable(gsi);
+	gsi_evt_ring_teardown(gsi);
+}
+
+/* Setup function for GSI.  GSI firmware must be loaded and initialized */
+int gsi_setup(struct gsi *gsi, bool db_enable)
+{
+	u32 val;
+
+	/* Here is where we first touch the GSI hardware */
+	val = ioread32(gsi->virt + GSI_GSI_STATUS_OFFSET);
+	if (!(val & ENABLED_FMASK)) {
+		dev_err(gsi->dev, "GSI has not been enabled\n");
+		return -EIO;
+	}
+
+	val = ioread32(gsi->virt + GSI_GSI_HW_PARAM_2_OFFSET);
+
+	gsi->channel_count = u32_get_bits(val, NUM_CH_PER_EE_FMASK);
+	if (!gsi->channel_count) {
+		dev_err(gsi->dev, "GSI reports zero channels supported\n");
+		return -EINVAL;
+	}
+	if (gsi->channel_count > GSI_CHANNEL_COUNT_MAX) {
+		dev_warn(gsi->dev,
+			"limiting to %u channels (hardware supports %u)\n",
+			 GSI_CHANNEL_COUNT_MAX, gsi->channel_count);
+		gsi->channel_count = GSI_CHANNEL_COUNT_MAX;
+	}
+
+	gsi->evt_ring_count = u32_get_bits(val, NUM_EV_PER_EE_FMASK);
+	if (!gsi->evt_ring_count) {
+		dev_err(gsi->dev, "GSI reports zero event rings supported\n");
+		return -EINVAL;
+	}
+	if (gsi->evt_ring_count > GSI_EVT_RING_COUNT_MAX) {
+		dev_warn(gsi->dev,
+			"limiting to %u event rings (hardware supports %u)\n",
+			 GSI_EVT_RING_COUNT_MAX, gsi->evt_ring_count);
+		gsi->evt_ring_count = GSI_EVT_RING_COUNT_MAX;
+	}
+
+	/* Initialize the error log */
+	iowrite32(0, gsi->virt + GSI_ERROR_LOG_OFFSET);
+
+	/* Writing 1 indicates IRQ interrupts; 0 would be MSI */
+	iowrite32(1, gsi->virt + GSI_CNTXT_INTSET_OFFSET);
+
+	return gsi_channel_setup(gsi, db_enable);
+}
+
+/* Inverse of gsi_setup() */
+void gsi_teardown(struct gsi *gsi)
+{
+	gsi_channel_teardown(gsi);
+}
+
+/* Initialize a channel's event ring */
+static int gsi_channel_evt_ring_init(struct gsi_channel *channel)
+{
+	struct gsi *gsi = channel->gsi;
+	struct gsi_evt_ring *evt_ring;
+	int ret;
+
+	ret = gsi_evt_ring_id_alloc(gsi);
+	if (ret < 0)
+		return ret;
+	channel->evt_ring_id = ret;
+
+	evt_ring = &gsi->evt_ring[channel->evt_ring_id];
+	evt_ring->channel = channel;
+
+	ret = gsi_ring_alloc(gsi, &evt_ring->ring, channel->event_count);
+	if (!ret)
+		return 0;	/* Success! */
+
+	dev_err(gsi->dev, "error %d allocating channel %u event ring\n",
+		ret, gsi_channel_id(channel));
+
+	gsi_evt_ring_id_free(gsi, channel->evt_ring_id);
+
+	return ret;
+}
+
+/* Inverse of gsi_channel_evt_ring_init() */
+static void gsi_channel_evt_ring_exit(struct gsi_channel *channel)
+{
+	u32 evt_ring_id = channel->evt_ring_id;
+	struct gsi *gsi = channel->gsi;
+	struct gsi_evt_ring *evt_ring;
+
+	evt_ring = &gsi->evt_ring[evt_ring_id];
+	gsi_ring_free(gsi, &evt_ring->ring);
+	gsi_evt_ring_id_free(gsi, evt_ring_id);
+}
+
+/* Init function for event rings */
+static void gsi_evt_ring_init(struct gsi *gsi)
+{
+	u32 evt_ring_id = 0;
+
+	gsi->event_bitmap = gsi_event_bitmap_init(GSI_EVT_RING_COUNT_MAX);
+	gsi->event_enable_bitmap = 0;
+	do
+		init_completion(&gsi->evt_ring[evt_ring_id].completion);
+	while (++evt_ring_id < GSI_EVT_RING_COUNT_MAX);
+}
+
+/* Inverse of gsi_evt_ring_init() */
+static void gsi_evt_ring_exit(struct gsi *gsi)
+{
+	/* Nothing to do */
+}
+
+static bool gsi_channel_data_valid(struct gsi *gsi,
+				   const struct ipa_gsi_endpoint_data *data)
+{
+#ifdef IPA_VALIDATION
+	u32 channel_id = data->channel_id;
+	struct device *dev = gsi->dev;
+
+	/* Make sure channel ids are in the range driver supports */
+	if (channel_id >= GSI_CHANNEL_COUNT_MAX) {
+		dev_err(dev, "bad channel id %u (must be less than %u)\n",
+			channel_id, GSI_CHANNEL_COUNT_MAX);
+		return false;
+	}
+
+	if (data->ee_id != GSI_EE_AP && data->ee_id != GSI_EE_MODEM) {
+		dev_err(dev, "bad EE id %u (AP or modem)\n", data->ee_id);
+		return false;
+	}
+
+	if (!data->channel.tlv_count ||
+	    data->channel.tlv_count > GSI_TLV_MAX) {
+		dev_err(dev, "channel %u bad tlv_count %u (must be 1..%u)\n",
+			channel_id, data->channel.tlv_count, GSI_TLV_MAX);
+		return false;
+	}
+
+	/* We have to allow at least one maximally-sized transaction to
+	 * be outstanding (which would use tlv_count TREs).  Given how
+	 * gsi_channel_tre_max() is computed, tre_count has to be almost
+	 * twice the TLV FIFO size to satisfy this requirement.
+	 */
+	if (data->channel.tre_count < 2 * data->channel.tlv_count - 1) {
+		dev_err(dev, "channel %u TLV count %u exceeds TRE count %u\n",
+			channel_id, data->channel.tlv_count,
+			data->channel.tre_count);
+		return false;
+	}
+
+	if (!is_power_of_2(data->channel.tre_count)) {
+		dev_err(dev, "channel %u bad tre_count %u (not power of 2)\n",
+			channel_id, data->channel.tre_count);
+		return false;
+	}
+
+	if (!is_power_of_2(data->channel.event_count)) {
+		dev_err(dev, "channel %u bad event_count %u (not power of 2)\n",
+			channel_id, data->channel.event_count);
+		return false;
+	}
+#endif /* IPA_VALIDATION */
+
+	return true;
+}
+
+/* Init function for a single channel */
+static int gsi_channel_init_one(struct gsi *gsi,
+				const struct ipa_gsi_endpoint_data *data,
+				bool command, bool prefetch)
+{
+	struct gsi_channel *channel;
+	u32 tre_count;
+	int ret;
+
+	if (!gsi_channel_data_valid(gsi, data))
+		return -EINVAL;
+
+	/* Worst case we need an event for every outstanding TRE */
+	if (data->channel.tre_count > data->channel.event_count) {
+		dev_warn(gsi->dev, "channel %u limited to %u TREs\n",
+			data->channel_id, data->channel.tre_count);
+		tre_count = data->channel.event_count;
+	} else {
+		tre_count = data->channel.tre_count;
+	}
+
+	channel = &gsi->channel[data->channel_id];
+	memset(channel, 0, sizeof(*channel));
+
+	channel->gsi = gsi;
+	channel->toward_ipa = data->toward_ipa;
+	channel->command = command;
+	channel->use_prefetch = command && prefetch;
+	channel->tlv_count = data->channel.tlv_count;
+	channel->tre_count = tre_count;
+	channel->event_count = data->channel.event_count;
+	init_completion(&channel->completion);
+
+	ret = gsi_channel_evt_ring_init(channel);
+	if (ret)
+		goto err_clear_gsi;
+
+	ret = gsi_ring_alloc(gsi, &channel->tre_ring, data->channel.tre_count);
+	if (ret) {
+		dev_err(gsi->dev, "error %d allocating channel %u ring\n",
+			ret, data->channel_id);
+		goto err_channel_evt_ring_exit;
+	}
+
+	ret = gsi_channel_trans_init(gsi, data->channel_id);
+	if (ret)
+		goto err_ring_free;
+
+	if (command) {
+		u32 tre_max = gsi_channel_tre_max(gsi, data->channel_id);
+
+		ret = ipa_cmd_pool_init(channel, tre_max);
+	}
+	if (!ret)
+		return 0;	/* Success! */
+
+	gsi_channel_trans_exit(channel);
+err_ring_free:
+	gsi_ring_free(gsi, &channel->tre_ring);
+err_channel_evt_ring_exit:
+	gsi_channel_evt_ring_exit(channel);
+err_clear_gsi:
+	channel->gsi = NULL;	/* Mark it not (fully) initialized */
+
+	return ret;
+}
+
+/* Inverse of gsi_channel_init_one() */
+static void gsi_channel_exit_one(struct gsi_channel *channel)
+{
+	if (!channel->gsi)
+		return;		/* Ignore uninitialized channels */
+
+	if (channel->command)
+		ipa_cmd_pool_exit(channel);
+	gsi_channel_trans_exit(channel);
+	gsi_ring_free(channel->gsi, &channel->tre_ring);
+	gsi_channel_evt_ring_exit(channel);
+}
+
+/* Init function for channels */
+static int gsi_channel_init(struct gsi *gsi, bool prefetch, u32 count,
+			    const struct ipa_gsi_endpoint_data *data,
+			    bool modem_alloc)
+{
+	int ret = 0;
+	u32 i;
+
+	gsi_evt_ring_init(gsi);
+
+	/* The endpoint data array is indexed by endpoint name */
+	for (i = 0; i < count; i++) {
+		bool command = i == IPA_ENDPOINT_AP_COMMAND_TX;
+
+		if (ipa_gsi_endpoint_data_empty(&data[i]))
+			continue;	/* Skip over empty slots */
+
+		/* Mark modem channels to be allocated (hardware workaround) */
+		if (data[i].ee_id == GSI_EE_MODEM) {
+			if (modem_alloc)
+				gsi->modem_channel_bitmap |=
+						BIT(data[i].channel_id);
+			continue;
+		}
+
+		ret = gsi_channel_init_one(gsi, &data[i], command, prefetch);
+		if (ret)
+			goto err_unwind;
+	}
+
+	return ret;
+
+err_unwind:
+	while (i--) {
+		if (ipa_gsi_endpoint_data_empty(&data[i]))
+			continue;
+		if (modem_alloc && data[i].ee_id == GSI_EE_MODEM) {
+			gsi->modem_channel_bitmap &= ~BIT(data[i].channel_id);
+			continue;
+		}
+		gsi_channel_exit_one(&gsi->channel[data->channel_id]);
+	}
+	gsi_evt_ring_exit(gsi);
+
+	return ret;
+}
+
+/* Inverse of gsi_channel_init() */
+static void gsi_channel_exit(struct gsi *gsi)
+{
+	u32 channel_id = GSI_CHANNEL_COUNT_MAX - 1;
+
+	do
+		gsi_channel_exit_one(&gsi->channel[channel_id]);
+	while (channel_id--);
+	gsi->modem_channel_bitmap = 0;
+
+	gsi_evt_ring_exit(gsi);
+}
+
+/* Init function for GSI.  GSI hardware does not need to be "ready" */
+int gsi_init(struct gsi *gsi, struct platform_device *pdev, bool prefetch,
+	     u32 count, const struct ipa_gsi_endpoint_data *data,
+	     bool modem_alloc)
+{
+	struct resource *res;
+	resource_size_t size;
+	unsigned int irq;
+	int ret;
+
+	gsi_validate_build();
+
+	gsi->dev = &pdev->dev;
+
+	/* The GSI layer performs NAPI on all endpoints.  NAPI requires a
+	 * network device structure, but the GSI layer does not have one,
+	 * so we must create a dummy network device for this purpose.
+	 */
+	init_dummy_netdev(&gsi->dummy_dev);
+
+	/* Get the GSI IRQ and request for it to wake the system */
+	ret = platform_get_irq_byname(pdev, "gsi");
+	if (ret <= 0) {
+		dev_err(gsi->dev,
+			"DT error %d getting \"gsi\" IRQ property\n", ret);
+		return ret ? : -EINVAL;
+	}
+	irq = ret;
+
+	ret = request_irq(irq, gsi_isr, 0, "gsi", gsi);
+	if (ret) {
+		dev_err(gsi->dev, "error %d requesting \"gsi\" IRQ\n", ret);
+		return ret;
+	}
+	gsi->irq = irq;
+
+	ret = enable_irq_wake(gsi->irq);
+	if (ret)
+		dev_warn(gsi->dev, "error %d enabling gsi wake irq\n", ret);
+	gsi->irq_wake_enabled = !ret;
+
+	/* Get GSI memory range and map it */
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "gsi");
+	if (!res) {
+		dev_err(gsi->dev,
+			"DT error getting \"gsi\" memory property\n");
+		ret = -ENODEV;
+		goto err_disable_irq_wake;
+	}
+
+	size = resource_size(res);
+	if (res->start > U32_MAX || size > U32_MAX - res->start) {
+		dev_err(gsi->dev, "DT memory resource \"gsi\" out of range\n");
+		ret = -EINVAL;
+		goto err_disable_irq_wake;
+	}
+
+	gsi->virt = ioremap(res->start, size);
+	if (!gsi->virt) {
+		dev_err(gsi->dev, "unable to remap \"gsi\" memory\n");
+		ret = -ENOMEM;
+		goto err_disable_irq_wake;
+	}
+
+	ret = gsi_channel_init(gsi, prefetch, count, data, modem_alloc);
+	if (ret)
+		goto err_iounmap;
+
+	mutex_init(&gsi->mutex);
+	init_completion(&gsi->completion);
+
+	return 0;
+
+err_iounmap:
+	iounmap(gsi->virt);
+err_disable_irq_wake:
+	if (gsi->irq_wake_enabled)
+		(void)disable_irq_wake(gsi->irq);
+	free_irq(gsi->irq, gsi);
+
+	return ret;
+}
+
+/* Inverse of gsi_init() */
+void gsi_exit(struct gsi *gsi)
+{
+	mutex_destroy(&gsi->mutex);
+	gsi_channel_exit(gsi);
+	if (gsi->irq_wake_enabled)
+		(void)disable_irq_wake(gsi->irq);
+	free_irq(gsi->irq, gsi);
+	iounmap(gsi->virt);
+}
+
+/* The maximum number of outstanding TREs on a channel.  This limits
+ * a channel's maximum number of transactions outstanding (worst case
+ * is one TRE per transaction).
+ *
+ * The absolute limit is the number of TREs in the channel's TRE ring,
+ * and in theory we should be able use all of them.  But in practice,
+ * doing that led to the hardware reporting exhaustion of event ring
+ * slots for writing completion information.  So the hardware limit
+ * would be (tre_count - 1).
+ *
+ * We reduce it a bit further though.  Transaction resource pools are
+ * sized to be a little larger than this maximum, to allow resource
+ * allocations to always be contiguous.  The number of entries in a
+ * TRE ring buffer is a power of 2, and the extra resources in a pool
+ * tends to nearly double the memory allocated for it.  Reducing the
+ * maximum number of outstanding TREs allows the number of entries in
+ * a pool to avoid crossing that power-of-2 boundary, and this can
+ * substantially reduce pool memory requirements.  The number we
+ * reduce it by matches the number added in gsi_trans_pool_init().
+ */
+u32 gsi_channel_tre_max(struct gsi *gsi, u32 channel_id)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+
+	/* Hardware limit is channel->tre_count - 1 */
+	return channel->tre_count - (channel->tlv_count - 1);
+}
+
+/* Returns the maximum number of TREs in a single transaction for a channel */
+u32 gsi_channel_trans_tre_max(struct gsi *gsi, u32 channel_id)
+{
+	struct gsi_channel *channel = &gsi->channel[channel_id];
+
+	return channel->tlv_count;
+}