diff --git a/Documentation/ABI/testing/sysfs-bus-intel_th-devices-msc b/Documentation/ABI/testing/sysfs-bus-intel_th-devices-msc
new file mode 100644
index 0000000000000000000000000000000000000000..b940c5d91cf7a37e32cbf92690cf830712c671de
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-bus-intel_th-devices-msc
@@ -0,0 +1,33 @@
+What:		/sys/bus/intel_th/devices/<intel_th_id>-msc<msc-id>/wrap
+Date:		June 2015
+KernelVersion:	4.3
+Contact:	Alexander Shishkin <alexander.shishkin@linux.intel.com>
+Description:	(RW) Configure MSC buffer wrapping. 1 == wrapping enabled.
+
+What:		/sys/bus/intel_th/devices/<intel_th_id>-msc<msc-id>/mode
+Date:		June 2015
+KernelVersion:	4.3
+Contact:	Alexander Shishkin <alexander.shishkin@linux.intel.com>
+Description:	(RW) Configure MSC operating mode:
+		  - "single", for contiguous buffer mode (high-order alloc);
+		  - "multi", for multiblock mode;
+		  - "ExI", for DCI handler mode;
+		  - "debug", for debug mode.
+		If operating mode changes, existing buffer is deallocated,
+		provided there are no active users and tracing is not enabled,
+		otherwise the write will fail.
+
+What:		/sys/bus/intel_th/devices/<intel_th_id>-msc<msc-id>/nr_pages
+Date:		June 2015
+KernelVersion:	4.3
+Contact:	Alexander Shishkin <alexander.shishkin@linux.intel.com>
+Description:	(RW) Configure MSC buffer size for "single" or "multi" modes.
+		In single mode, this is a single number of pages, has to be
+		power of 2. In multiblock mode, this is a comma-separated list
+		of numbers of pages for each window to be allocated. Number of
+		windows is not limited.
+		Writing to this file deallocates existing buffer (provided
+		there are no active users and tracing is not enabled) and then
+		allocates a new one.
+
+
diff --git a/drivers/hwtracing/intel_th/Kconfig b/drivers/hwtracing/intel_th/Kconfig
index b0855026c9f34aaa9dd05c67a70fdcc2985f4d2c..d9a1217e27bf7600ad27ab264a76cce0b1322383 100644
--- a/drivers/hwtracing/intel_th/Kconfig
+++ b/drivers/hwtracing/intel_th/Kconfig
@@ -44,6 +44,16 @@ config INTEL_TH_STH
 
 	  Say Y here to enable STH subdevice of Intel(R) Trace Hub.
 
+config INTEL_TH_MSU
+	tristate "Intel(R) Trace Hub Memory Storage Unit"
+	help
+	  Memory Storage Unit (MSU) trace output device enables
+	  storing STP traces to system memory. It supports single
+	  and multiblock modes of operation and provides read()
+	  and mmap() access to the collected data.
+
+	  Say Y here to enable MSU output device for Intel TH.
+
 config INTEL_TH_DEBUG
 	bool "Intel(R) Trace Hub debugging"
 	depends on DEBUG_FS
diff --git a/drivers/hwtracing/intel_th/Makefile b/drivers/hwtracing/intel_th/Makefile
index 8c45e9fa6fc02e84ef9323505e80d78cb559b711..4636757df2a1e20a6c790942c252d421af04f0eb 100644
--- a/drivers/hwtracing/intel_th/Makefile
+++ b/drivers/hwtracing/intel_th/Makefile
@@ -10,3 +10,6 @@ intel_th_gth-y			:= gth.o
 
 obj-$(CONFIG_INTEL_TH_STH)	+= intel_th_sth.o
 intel_th_sth-y			:= sth.o
+
+obj-$(CONFIG_INTEL_TH_MSU)	+= intel_th_msu.o
+intel_th_msu-y			:= msu.o
diff --git a/drivers/hwtracing/intel_th/msu.c b/drivers/hwtracing/intel_th/msu.c
new file mode 100644
index 0000000000000000000000000000000000000000..99f7662e27f15e3b20d9ab392014baa19db95c91
--- /dev/null
+++ b/drivers/hwtracing/intel_th/msu.c
@@ -0,0 +1,1509 @@
+/*
+ * Intel(R) Trace Hub Memory Storage Unit
+ *
+ * Copyright (C) 2014-2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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.
+ */
+
+#define pr_fmt(fmt)	KBUILD_MODNAME ": " fmt
+
+#include <linux/types.h>
+#include <linux/module.h>
+#include <linux/device.h>
+#include <linux/uaccess.h>
+#include <linux/sizes.h>
+#include <linux/printk.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+#include <linux/fs.h>
+#include <linux/io.h>
+#include <linux/dma-mapping.h>
+
+#include <asm/cacheflush.h>
+
+#include "intel_th.h"
+#include "msu.h"
+
+#define msc_dev(x) (&(x)->thdev->dev)
+
+/**
+ * struct msc_block - multiblock mode block descriptor
+ * @bdesc:	pointer to hardware descriptor (beginning of the block)
+ * @addr:	physical address of the block
+ */
+struct msc_block {
+	struct msc_block_desc	*bdesc;
+	dma_addr_t		addr;
+};
+
+/**
+ * struct msc_window - multiblock mode window descriptor
+ * @entry:	window list linkage (msc::win_list)
+ * @pgoff:	page offset into the buffer that this window starts at
+ * @nr_blocks:	number of blocks (pages) in this window
+ * @block:	array of block descriptors
+ */
+struct msc_window {
+	struct list_head	entry;
+	unsigned long		pgoff;
+	unsigned int		nr_blocks;
+	struct msc		*msc;
+	struct msc_block	block[0];
+};
+
+/**
+ * struct msc_iter - iterator for msc buffer
+ * @entry:		msc::iter_list linkage
+ * @msc:		pointer to the MSC device
+ * @start_win:		oldest window
+ * @win:		current window
+ * @offset:		current logical offset into the buffer
+ * @start_block:	oldest block in the window
+ * @block:		block number in the window
+ * @block_off:		offset into current block
+ * @wrap_count:		block wrapping handling
+ * @eof:		end of buffer reached
+ */
+struct msc_iter {
+	struct list_head	entry;
+	struct msc		*msc;
+	struct msc_window	*start_win;
+	struct msc_window	*win;
+	unsigned long		offset;
+	int			start_block;
+	int			block;
+	unsigned int		block_off;
+	unsigned int		wrap_count;
+	unsigned int		eof;
+};
+
+/**
+ * struct msc - MSC device representation
+ * @reg_base:		register window base address
+ * @thdev:		intel_th_device pointer
+ * @win_list:		list of windows in multiblock mode
+ * @nr_pages:		total number of pages allocated for this buffer
+ * @single_sz:		amount of data in single mode
+ * @single_wrap:	single mode wrap occurred
+ * @base:		buffer's base pointer
+ * @base_addr:		buffer's base address
+ * @user_count:		number of users of the buffer
+ * @mmap_count:		number of mappings
+ * @buf_mutex:		mutex to serialize access to buffer-related bits
+
+ * @enabled:		MSC is enabled
+ * @wrap:		wrapping is enabled
+ * @mode:		MSC operating mode
+ * @burst_len:		write burst length
+ * @index:		number of this MSC in the MSU
+ */
+struct msc {
+	void __iomem		*reg_base;
+	struct intel_th_device	*thdev;
+
+	struct list_head	win_list;
+	unsigned long		nr_pages;
+	unsigned long		single_sz;
+	unsigned int		single_wrap : 1;
+	void			*base;
+	dma_addr_t		base_addr;
+
+	/* <0: no buffer, 0: no users, >0: active users */
+	atomic_t		user_count;
+
+	atomic_t		mmap_count;
+	struct mutex		buf_mutex;
+
+	struct mutex		iter_mutex;
+	struct list_head	iter_list;
+
+	/* config */
+	unsigned int		enabled : 1,
+				wrap	: 1;
+	unsigned int		mode;
+	unsigned int		burst_len;
+	unsigned int		index;
+};
+
+static inline bool msc_block_is_empty(struct msc_block_desc *bdesc)
+{
+	/* header hasn't been written */
+	if (!bdesc->valid_dw)
+		return true;
+
+	/* valid_dw includes the header */
+	if (!msc_data_sz(bdesc))
+		return true;
+
+	return false;
+}
+
+/**
+ * msc_oldest_window() - locate the window with oldest data
+ * @msc:	MSC device
+ *
+ * This should only be used in multiblock mode. Caller should hold the
+ * msc::user_count reference.
+ *
+ * Return:	the oldest window with valid data
+ */
+static struct msc_window *msc_oldest_window(struct msc *msc)
+{
+	struct msc_window *win;
+	u32 reg = ioread32(msc->reg_base + REG_MSU_MSC0NWSA);
+	unsigned long win_addr = (unsigned long)reg << PAGE_SHIFT;
+	unsigned int found = 0;
+
+	if (list_empty(&msc->win_list))
+		return NULL;
+
+	/*
+	 * we might need a radix tree for this, depending on how
+	 * many windows a typical user would allocate; ideally it's
+	 * something like 2, in which case we're good
+	 */
+	list_for_each_entry(win, &msc->win_list, entry) {
+		if (win->block[0].addr == win_addr)
+			found++;
+
+		/* skip the empty ones */
+		if (msc_block_is_empty(win->block[0].bdesc))
+			continue;
+
+		if (found)
+			return win;
+	}
+
+	return list_entry(msc->win_list.next, struct msc_window, entry);
+}
+
+/**
+ * msc_win_oldest_block() - locate the oldest block in a given window
+ * @win:	window to look at
+ *
+ * Return:	index of the block with the oldest data
+ */
+static unsigned int msc_win_oldest_block(struct msc_window *win)
+{
+	unsigned int blk;
+	struct msc_block_desc *bdesc = win->block[0].bdesc;
+
+	/* without wrapping, first block is the oldest */
+	if (!msc_block_wrapped(bdesc))
+		return 0;
+
+	/*
+	 * with wrapping, last written block contains both the newest and the
+	 * oldest data for this window.
+	 */
+	for (blk = 0; blk < win->nr_blocks; blk++) {
+		bdesc = win->block[blk].bdesc;
+
+		if (msc_block_last_written(bdesc))
+			return blk;
+	}
+
+	return 0;
+}
+
+/**
+ * msc_is_last_win() - check if a window is the last one for a given MSC
+ * @win:	window
+ * Return:	true if @win is the last window in MSC's multiblock buffer
+ */
+static inline bool msc_is_last_win(struct msc_window *win)
+{
+	return win->entry.next == &win->msc->win_list;
+}
+
+/**
+ * msc_next_window() - return next window in the multiblock buffer
+ * @win:	current window
+ *
+ * Return:	window following the current one
+ */
+static struct msc_window *msc_next_window(struct msc_window *win)
+{
+	if (msc_is_last_win(win))
+		return list_entry(win->msc->win_list.next, struct msc_window,
+				  entry);
+
+	return list_entry(win->entry.next, struct msc_window, entry);
+}
+
+static struct msc_block_desc *msc_iter_bdesc(struct msc_iter *iter)
+{
+	return iter->win->block[iter->block].bdesc;
+}
+
+static void msc_iter_init(struct msc_iter *iter)
+{
+	memset(iter, 0, sizeof(*iter));
+	iter->start_block = -1;
+	iter->block = -1;
+}
+
+static struct msc_iter *msc_iter_install(struct msc *msc)
+{
+	struct msc_iter *iter;
+
+	iter = kzalloc(sizeof(*iter), GFP_KERNEL);
+	if (!iter)
+		return NULL;
+
+	msc_iter_init(iter);
+	iter->msc = msc;
+
+	mutex_lock(&msc->iter_mutex);
+	list_add_tail(&iter->entry, &msc->iter_list);
+	mutex_unlock(&msc->iter_mutex);
+
+	return iter;
+}
+
+static void msc_iter_remove(struct msc_iter *iter, struct msc *msc)
+{
+	mutex_lock(&msc->iter_mutex);
+	list_del(&iter->entry);
+	mutex_unlock(&msc->iter_mutex);
+
+	kfree(iter);
+}
+
+static void msc_iter_block_start(struct msc_iter *iter)
+{
+	if (iter->start_block != -1)
+		return;
+
+	iter->start_block = msc_win_oldest_block(iter->win);
+	iter->block = iter->start_block;
+	iter->wrap_count = 0;
+
+	/*
+	 * start with the block with oldest data; if data has wrapped
+	 * in this window, it should be in this block
+	 */
+	if (msc_block_wrapped(msc_iter_bdesc(iter)))
+		iter->wrap_count = 2;
+
+}
+
+static int msc_iter_win_start(struct msc_iter *iter, struct msc *msc)
+{
+	/* already started, nothing to do */
+	if (iter->start_win)
+		return 0;
+
+	iter->start_win = msc_oldest_window(msc);
+	if (!iter->start_win)
+		return -EINVAL;
+
+	iter->win = iter->start_win;
+	iter->start_block = -1;
+
+	msc_iter_block_start(iter);
+
+	return 0;
+}
+
+static int msc_iter_win_advance(struct msc_iter *iter)
+{
+	iter->win = msc_next_window(iter->win);
+	iter->start_block = -1;
+
+	if (iter->win == iter->start_win) {
+		iter->eof++;
+		return 1;
+	}
+
+	msc_iter_block_start(iter);
+
+	return 0;
+}
+
+static int msc_iter_block_advance(struct msc_iter *iter)
+{
+	iter->block_off = 0;
+
+	/* wrapping */
+	if (iter->wrap_count && iter->block == iter->start_block) {
+		iter->wrap_count--;
+		if (!iter->wrap_count)
+			/* copied newest data from the wrapped block */
+			return msc_iter_win_advance(iter);
+	}
+
+	/* no wrapping, check for last written block */
+	if (!iter->wrap_count && msc_block_last_written(msc_iter_bdesc(iter)))
+		/* copied newest data for the window */
+		return msc_iter_win_advance(iter);
+
+	/* block advance */
+	if (++iter->block == iter->win->nr_blocks)
+		iter->block = 0;
+
+	/* no wrapping, sanity check in case there is no last written block */
+	if (!iter->wrap_count && iter->block == iter->start_block)
+		return msc_iter_win_advance(iter);
+
+	return 0;
+}
+
+/**
+ * msc_buffer_iterate() - go through multiblock buffer's data
+ * @iter:	iterator structure
+ * @size:	amount of data to scan
+ * @data:	callback's private data
+ * @fn:		iterator callback
+ *
+ * This will start at the window which will be written to next (containing
+ * the oldest data) and work its way to the current window, calling @fn
+ * for each chunk of data as it goes.
+ *
+ * Caller should have msc::user_count reference to make sure the buffer
+ * doesn't disappear from under us.
+ *
+ * Return:	amount of data actually scanned.
+ */
+static ssize_t
+msc_buffer_iterate(struct msc_iter *iter, size_t size, void *data,
+		   unsigned long (*fn)(void *, void *, size_t))
+{
+	struct msc *msc = iter->msc;
+	size_t len = size;
+	unsigned int advance;
+
+	if (iter->eof)
+		return 0;
+
+	/* start with the oldest window */
+	if (msc_iter_win_start(iter, msc))
+		return 0;
+
+	do {
+		unsigned long data_bytes = msc_data_sz(msc_iter_bdesc(iter));
+		void *src = (void *)msc_iter_bdesc(iter) + MSC_BDESC;
+		size_t tocopy = data_bytes, copied = 0;
+		size_t remaining = 0;
+
+		advance = 1;
+
+		/*
+		 * If block wrapping happened, we need to visit the last block
+		 * twice, because it contains both the oldest and the newest
+		 * data in this window.
+		 *
+		 * First time (wrap_count==2), in the very beginning, to collect
+		 * the oldest data, which is in the range
+		 * (data_bytes..DATA_IN_PAGE).
+		 *
+		 * Second time (wrap_count==1), it's just like any other block,
+		 * containing data in the range of [MSC_BDESC..data_bytes].
+		 */
+		if (iter->block == iter->start_block && iter->wrap_count) {
+			tocopy = DATA_IN_PAGE - data_bytes;
+			src += data_bytes;
+		}
+
+		if (!tocopy)
+			goto next_block;
+
+		tocopy -= iter->block_off;
+		src += iter->block_off;
+
+		if (len < tocopy) {
+			tocopy = len;
+			advance = 0;
+		}
+
+		remaining = fn(data, src, tocopy);
+
+		if (remaining)
+			advance = 0;
+
+		copied = tocopy - remaining;
+		len -= copied;
+		iter->block_off += copied;
+		iter->offset += copied;
+
+		if (!advance)
+			break;
+
+next_block:
+		if (msc_iter_block_advance(iter))
+			break;
+
+	} while (len);
+
+	return size - len;
+}
+
+/**
+ * msc_buffer_clear_hw_header() - clear hw header for multiblock
+ * @msc:	MSC device
+ */
+static void msc_buffer_clear_hw_header(struct msc *msc)
+{
+	struct msc_window *win;
+
+	mutex_lock(&msc->buf_mutex);
+	list_for_each_entry(win, &msc->win_list, entry) {
+		unsigned int blk;
+		size_t hw_sz = sizeof(struct msc_block_desc) -
+			offsetof(struct msc_block_desc, hw_tag);
+
+		for (blk = 0; blk < win->nr_blocks; blk++) {
+			struct msc_block_desc *bdesc = win->block[blk].bdesc;
+
+			memset(&bdesc->hw_tag, 0, hw_sz);
+		}
+	}
+	mutex_unlock(&msc->buf_mutex);
+}
+
+/**
+ * msc_configure() - set up MSC hardware
+ * @msc:	the MSC device to configure
+ *
+ * Program storage mode, wrapping, burst length and trace buffer address
+ * into a given MSC. If msc::enabled is set, enable the trace, too.
+ */
+static int msc_configure(struct msc *msc)
+{
+	u32 reg;
+
+	if (msc->mode > MSC_MODE_MULTI)
+		return -ENOTSUPP;
+
+	if (msc->mode == MSC_MODE_MULTI)
+		msc_buffer_clear_hw_header(msc);
+
+	reg = msc->base_addr >> PAGE_SHIFT;
+	iowrite32(reg, msc->reg_base + REG_MSU_MSC0BAR);
+
+	if (msc->mode == MSC_MODE_SINGLE) {
+		reg = msc->nr_pages;
+		iowrite32(reg, msc->reg_base + REG_MSU_MSC0SIZE);
+	}
+
+	reg = ioread32(msc->reg_base + REG_MSU_MSC0CTL);
+	reg &= ~(MSC_MODE | MSC_WRAPEN | MSC_EN | MSC_RD_HDR_OVRD);
+
+	reg |= msc->mode << __ffs(MSC_MODE);
+	reg |= msc->burst_len << __ffs(MSC_LEN);
+	/*if (msc->mode == MSC_MODE_MULTI)
+	  reg |= MSC_RD_HDR_OVRD; */
+	if (msc->wrap)
+		reg |= MSC_WRAPEN;
+	if (msc->enabled)
+		reg |= MSC_EN;
+
+	iowrite32(reg, msc->reg_base + REG_MSU_MSC0CTL);
+
+	if (msc->enabled) {
+		msc->thdev->output.multiblock = msc->mode == MSC_MODE_MULTI;
+		intel_th_trace_enable(msc->thdev);
+	}
+
+	return 0;
+}
+
+/**
+ * msc_disable() - disable MSC hardware
+ * @msc:	MSC device to disable
+ *
+ * If @msc is enabled, disable tracing on the switch and then disable MSC
+ * storage.
+ */
+static void msc_disable(struct msc *msc)
+{
+	unsigned long count;
+	u32 reg;
+
+	if (!msc->enabled)
+		return;
+
+	intel_th_trace_disable(msc->thdev);
+
+	for (reg = 0, count = MSC_PLE_WAITLOOP_DEPTH;
+	     count && !(reg & MSCSTS_PLE); count--) {
+		reg = ioread32(msc->reg_base + REG_MSU_MSC0STS);
+		cpu_relax();
+	}
+
+	if (!count)
+		dev_dbg(msc_dev(msc), "timeout waiting for MSC0 PLE\n");
+
+	if (msc->mode == MSC_MODE_SINGLE) {
+		msc->single_wrap = !!(reg & MSCSTS_WRAPSTAT);
+
+		reg = ioread32(msc->reg_base + REG_MSU_MSC0MWP);
+		msc->single_sz = reg & ((msc->nr_pages << PAGE_SHIFT) - 1);
+		dev_dbg(msc_dev(msc), "MSCnMWP: %08x/%08lx, wrap: %d\n",
+			reg, msc->single_sz, msc->single_wrap);
+	}
+
+	reg = ioread32(msc->reg_base + REG_MSU_MSC0CTL);
+	reg &= ~MSC_EN;
+	iowrite32(reg, msc->reg_base + REG_MSU_MSC0CTL);
+	msc->enabled = 0;
+
+	iowrite32(0, msc->reg_base + REG_MSU_MSC0BAR);
+	iowrite32(0, msc->reg_base + REG_MSU_MSC0SIZE);
+
+	dev_dbg(msc_dev(msc), "MSCnNWSA: %08x\n",
+		ioread32(msc->reg_base + REG_MSU_MSC0NWSA));
+
+	reg = ioread32(msc->reg_base + REG_MSU_MSC0STS);
+	dev_dbg(msc_dev(msc), "MSCnSTS: %08x\n", reg);
+}
+
+static int intel_th_msc_activate(struct intel_th_device *thdev)
+{
+	struct msc *msc = dev_get_drvdata(&thdev->dev);
+	int ret = 0;
+
+	if (!atomic_inc_unless_negative(&msc->user_count))
+		return -ENODEV;
+
+	mutex_lock(&msc->iter_mutex);
+	if (!list_empty(&msc->iter_list))
+		ret = -EBUSY;
+	mutex_unlock(&msc->iter_mutex);
+
+	if (ret) {
+		atomic_dec(&msc->user_count);
+		return ret;
+	}
+
+	msc->enabled = 1;
+
+	return msc_configure(msc);
+}
+
+static void intel_th_msc_deactivate(struct intel_th_device *thdev)
+{
+	struct msc *msc = dev_get_drvdata(&thdev->dev);
+
+	msc_disable(msc);
+
+	atomic_dec(&msc->user_count);
+}
+
+/**
+ * msc_buffer_contig_alloc() - allocate a contiguous buffer for SINGLE mode
+ * @msc:	MSC device
+ * @size:	allocation size in bytes
+ *
+ * This modifies msc::base, which requires msc::buf_mutex to serialize, so the
+ * caller is expected to hold it.
+ *
+ * Return:	0 on success, -errno otherwise.
+ */
+static int msc_buffer_contig_alloc(struct msc *msc, unsigned long size)
+{
+	unsigned int order = get_order(size);
+	struct page *page;
+
+	if (!size)
+		return 0;
+
+	page = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
+	if (!page)
+		return -ENOMEM;
+
+	split_page(page, order);
+	msc->nr_pages = size >> PAGE_SHIFT;
+	msc->base = page_address(page);
+	msc->base_addr = page_to_phys(page);
+
+	return 0;
+}
+
+/**
+ * msc_buffer_contig_free() - free a contiguous buffer
+ * @msc:	MSC configured in SINGLE mode
+ */
+static void msc_buffer_contig_free(struct msc *msc)
+{
+	unsigned long off;
+
+	for (off = 0; off < msc->nr_pages << PAGE_SHIFT; off += PAGE_SIZE) {
+		struct page *page = virt_to_page(msc->base + off);
+
+		page->mapping = NULL;
+		__free_page(page);
+	}
+
+	msc->nr_pages = 0;
+}
+
+/**
+ * msc_buffer_contig_get_page() - find a page at a given offset
+ * @msc:	MSC configured in SINGLE mode
+ * @pgoff:	page offset
+ *
+ * Return:	page, if @pgoff is within the range, NULL otherwise.
+ */
+static struct page *msc_buffer_contig_get_page(struct msc *msc,
+					       unsigned long pgoff)
+{
+	if (pgoff >= msc->nr_pages)
+		return NULL;
+
+	return virt_to_page(msc->base + (pgoff << PAGE_SHIFT));
+}
+
+/**
+ * msc_buffer_win_alloc() - alloc a window for a multiblock mode
+ * @msc:	MSC device
+ * @nr_blocks:	number of pages in this window
+ *
+ * This modifies msc::win_list and msc::base, which requires msc::buf_mutex
+ * to serialize, so the caller is expected to hold it.
+ *
+ * Return:	0 on success, -errno otherwise.
+ */
+static int msc_buffer_win_alloc(struct msc *msc, unsigned int nr_blocks)
+{
+	struct msc_window *win;
+	unsigned long size = PAGE_SIZE;
+	int i, ret = -ENOMEM;
+
+	if (!nr_blocks)
+		return 0;
+
+	win = kzalloc(offsetof(struct msc_window, block[nr_blocks]),
+		      GFP_KERNEL);
+	if (!win)
+		return -ENOMEM;
+
+	if (!list_empty(&msc->win_list)) {
+		struct msc_window *prev = list_entry(msc->win_list.prev,
+						     struct msc_window, entry);
+
+		win->pgoff = prev->pgoff + prev->nr_blocks;
+	}
+
+	for (i = 0; i < nr_blocks; i++) {
+		win->block[i].bdesc = dma_alloc_coherent(msc_dev(msc), size,
+							 &win->block[i].addr,
+							 GFP_KERNEL);
+
+#ifdef CONFIG_X86
+		/* Set the page as uncached */
+		set_memory_uc((unsigned long)win->block[i].bdesc, 1);
+#endif
+
+		if (!win->block[i].bdesc)
+			goto err_nomem;
+	}
+
+	win->msc = msc;
+	win->nr_blocks = nr_blocks;
+
+	if (list_empty(&msc->win_list)) {
+		msc->base = win->block[0].bdesc;
+		msc->base_addr = win->block[0].addr;
+	}
+
+	list_add_tail(&win->entry, &msc->win_list);
+	msc->nr_pages += nr_blocks;
+
+	return 0;
+
+err_nomem:
+	for (i--; i >= 0; i--) {
+#ifdef CONFIG_X86
+		/* Reset the page to write-back before releasing */
+		set_memory_wb((unsigned long)win->block[i].bdesc, 1);
+#endif
+		dma_free_coherent(msc_dev(msc), size, win->block[i].bdesc,
+				  win->block[i].addr);
+	}
+	kfree(win);
+
+	return ret;
+}
+
+/**
+ * msc_buffer_win_free() - free a window from MSC's window list
+ * @msc:	MSC device
+ * @win:	window to free
+ *
+ * This modifies msc::win_list and msc::base, which requires msc::buf_mutex
+ * to serialize, so the caller is expected to hold it.
+ */
+static void msc_buffer_win_free(struct msc *msc, struct msc_window *win)
+{
+	int i;
+
+	msc->nr_pages -= win->nr_blocks;
+
+	list_del(&win->entry);
+	if (list_empty(&msc->win_list)) {
+		msc->base = NULL;
+		msc->base_addr = 0;
+	}
+
+	for (i = 0; i < win->nr_blocks; i++) {
+		struct page *page = virt_to_page(win->block[i].bdesc);
+
+		page->mapping = NULL;
+#ifdef CONFIG_X86
+		/* Reset the page to write-back before releasing */
+		set_memory_wb((unsigned long)win->block[i].bdesc, 1);
+#endif
+		dma_free_coherent(msc_dev(win->msc), PAGE_SIZE,
+				  win->block[i].bdesc, win->block[i].addr);
+	}
+
+	kfree(win);
+}
+
+/**
+ * msc_buffer_relink() - set up block descriptors for multiblock mode
+ * @msc:	MSC device
+ *
+ * This traverses msc::win_list, which requires msc::buf_mutex to serialize,
+ * so the caller is expected to hold it.
+ */
+static void msc_buffer_relink(struct msc *msc)
+{
+	struct msc_window *win, *next_win;
+
+	/* call with msc::mutex locked */
+	list_for_each_entry(win, &msc->win_list, entry) {
+		unsigned int blk;
+		u32 sw_tag = 0;
+
+		/*
+		 * Last window's next_win should point to the first window
+		 * and MSC_SW_TAG_LASTWIN should be set.
+		 */
+		if (msc_is_last_win(win)) {
+			sw_tag |= MSC_SW_TAG_LASTWIN;
+			next_win = list_entry(msc->win_list.next,
+					      struct msc_window, entry);
+		} else {
+			next_win = list_entry(win->entry.next,
+					      struct msc_window, entry);
+		}
+
+		for (blk = 0; blk < win->nr_blocks; blk++) {
+			struct msc_block_desc *bdesc = win->block[blk].bdesc;
+
+			memset(bdesc, 0, sizeof(*bdesc));
+
+			bdesc->next_win = next_win->block[0].addr >> PAGE_SHIFT;
+
+			/*
+			 * Similarly to last window, last block should point
+			 * to the first one.
+			 */
+			if (blk == win->nr_blocks - 1) {
+				sw_tag |= MSC_SW_TAG_LASTBLK;
+				bdesc->next_blk =
+					win->block[0].addr >> PAGE_SHIFT;
+			} else {
+				bdesc->next_blk =
+					win->block[blk + 1].addr >> PAGE_SHIFT;
+			}
+
+			bdesc->sw_tag = sw_tag;
+			bdesc->block_sz = PAGE_SIZE / 64;
+		}
+	}
+
+	/*
+	 * Make the above writes globally visible before tracing is
+	 * enabled to make sure hardware sees them coherently.
+	 */
+	wmb();
+}
+
+static void msc_buffer_multi_free(struct msc *msc)
+{
+	struct msc_window *win, *iter;
+
+	list_for_each_entry_safe(win, iter, &msc->win_list, entry)
+		msc_buffer_win_free(msc, win);
+}
+
+static int msc_buffer_multi_alloc(struct msc *msc, unsigned long *nr_pages,
+				  unsigned int nr_wins)
+{
+	int ret, i;
+
+	for (i = 0; i < nr_wins; i++) {
+		ret = msc_buffer_win_alloc(msc, nr_pages[i]);
+		if (ret) {
+			msc_buffer_multi_free(msc);
+			return ret;
+		}
+	}
+
+	msc_buffer_relink(msc);
+
+	return 0;
+}
+
+/**
+ * msc_buffer_free() - free buffers for MSC
+ * @msc:	MSC device
+ *
+ * Free MSC's storage buffers.
+ *
+ * This modifies msc::win_list and msc::base, which requires msc::buf_mutex to
+ * serialize, so the caller is expected to hold it.
+ */
+static void msc_buffer_free(struct msc *msc)
+{
+	if (msc->mode == MSC_MODE_SINGLE)
+		msc_buffer_contig_free(msc);
+	else if (msc->mode == MSC_MODE_MULTI)
+		msc_buffer_multi_free(msc);
+}
+
+/**
+ * msc_buffer_alloc() - allocate a buffer for MSC
+ * @msc:	MSC device
+ * @size:	allocation size in bytes
+ *
+ * Allocate a storage buffer for MSC, depending on the msc::mode, it will be
+ * either done via msc_buffer_contig_alloc() for SINGLE operation mode or
+ * msc_buffer_win_alloc() for multiblock operation. The latter allocates one
+ * window per invocation, so in multiblock mode this can be called multiple
+ * times for the same MSC to allocate multiple windows.
+ *
+ * This modifies msc::win_list and msc::base, which requires msc::buf_mutex
+ * to serialize, so the caller is expected to hold it.
+ *
+ * Return:	0 on success, -errno otherwise.
+ */
+static int msc_buffer_alloc(struct msc *msc, unsigned long *nr_pages,
+			    unsigned int nr_wins)
+{
+	int ret;
+
+	/* -1: buffer not allocated */
+	if (atomic_read(&msc->user_count) != -1)
+		return -EBUSY;
+
+	if (msc->mode == MSC_MODE_SINGLE) {
+		if (nr_wins != 1)
+			return -EINVAL;
+
+		ret = msc_buffer_contig_alloc(msc, nr_pages[0] << PAGE_SHIFT);
+	} else if (msc->mode == MSC_MODE_MULTI) {
+		ret = msc_buffer_multi_alloc(msc, nr_pages, nr_wins);
+	} else {
+		ret = -ENOTSUPP;
+	}
+
+	if (!ret) {
+		/* allocation should be visible before the counter goes to 0 */
+		smp_mb__before_atomic();
+
+		if (WARN_ON_ONCE(atomic_cmpxchg(&msc->user_count, -1, 0) != -1))
+			return -EINVAL;
+	}
+
+	return ret;
+}
+
+/**
+ * msc_buffer_unlocked_free_unless_used() - free a buffer unless it's in use
+ * @msc:	MSC device
+ *
+ * This will free MSC buffer unless it is in use or there is no allocated
+ * buffer.
+ * Caller needs to hold msc::buf_mutex.
+ *
+ * Return:	0 on successful deallocation or if there was no buffer to
+ *		deallocate, -EBUSY if there are active users.
+ */
+static int msc_buffer_unlocked_free_unless_used(struct msc *msc)
+{
+	int count, ret = 0;
+
+	count = atomic_cmpxchg(&msc->user_count, 0, -1);
+
+	/* > 0: buffer is allocated and has users */
+	if (count > 0)
+		ret = -EBUSY;
+	/* 0: buffer is allocated, no users */
+	else if (!count)
+		msc_buffer_free(msc);
+	/* < 0: no buffer, nothing to do */
+
+	return ret;
+}
+
+/**
+ * msc_buffer_free_unless_used() - free a buffer unless it's in use
+ * @msc:	MSC device
+ *
+ * This is a locked version of msc_buffer_unlocked_free_unless_used().
+ */
+static int msc_buffer_free_unless_used(struct msc *msc)
+{
+	int ret;
+
+	mutex_lock(&msc->buf_mutex);
+	ret = msc_buffer_unlocked_free_unless_used(msc);
+	mutex_unlock(&msc->buf_mutex);
+
+	return ret;
+}
+
+/**
+ * msc_buffer_get_page() - get MSC buffer page at a given offset
+ * @msc:	MSC device
+ * @pgoff:	page offset into the storage buffer
+ *
+ * This traverses msc::win_list, so holding msc::buf_mutex is expected from
+ * the caller.
+ *
+ * Return:	page if @pgoff corresponds to a valid buffer page or NULL.
+ */
+static struct page *msc_buffer_get_page(struct msc *msc, unsigned long pgoff)
+{
+	struct msc_window *win;
+
+	if (msc->mode == MSC_MODE_SINGLE)
+		return msc_buffer_contig_get_page(msc, pgoff);
+
+	list_for_each_entry(win, &msc->win_list, entry)
+		if (pgoff >= win->pgoff && pgoff < win->pgoff + win->nr_blocks)
+			goto found;
+
+	return NULL;
+
+found:
+	pgoff -= win->pgoff;
+	return virt_to_page(win->block[pgoff].bdesc);
+}
+
+/**
+ * struct msc_win_to_user_struct - data for copy_to_user() callback
+ * @buf:	userspace buffer to copy data to
+ * @offset:	running offset
+ */
+struct msc_win_to_user_struct {
+	char __user	*buf;
+	unsigned long	offset;
+};
+
+/**
+ * msc_win_to_user() - iterator for msc_buffer_iterate() to copy data to user
+ * @data:	callback's private data
+ * @src:	source buffer
+ * @len:	amount of data to copy from the source buffer
+ */
+static unsigned long msc_win_to_user(void *data, void *src, size_t len)
+{
+	struct msc_win_to_user_struct *u = data;
+	unsigned long ret;
+
+	ret = copy_to_user(u->buf + u->offset, src, len);
+	u->offset += len - ret;
+
+	return ret;
+}
+
+
+/*
+ * file operations' callbacks
+ */
+
+static int intel_th_msc_open(struct inode *inode, struct file *file)
+{
+	struct intel_th_device *thdev = file->private_data;
+	struct msc *msc = dev_get_drvdata(&thdev->dev);
+	struct msc_iter *iter;
+
+	if (!capable(CAP_SYS_RAWIO))
+		return -EPERM;
+
+	iter = msc_iter_install(msc);
+	if (!iter)
+		return -ENOMEM;
+
+	file->private_data = iter;
+
+	return nonseekable_open(inode, file);
+}
+
+static int intel_th_msc_release(struct inode *inode, struct file *file)
+{
+	struct msc_iter *iter = file->private_data;
+	struct msc *msc = iter->msc;
+
+	msc_iter_remove(iter, msc);
+
+	return 0;
+}
+
+static ssize_t
+msc_single_to_user(struct msc *msc, char __user *buf, loff_t off, size_t len)
+{
+	size_t size = msc->nr_pages << PAGE_SHIFT, rem = len;
+	unsigned long start = off, tocopy = 0;
+
+	if (msc->single_wrap) {
+		start += msc->single_sz;
+		if (start < size) {
+			tocopy = min(rem, size - start);
+			if (copy_to_user(buf, msc->base + start, tocopy))
+				return -EFAULT;
+
+			buf += tocopy;
+			rem -= tocopy;
+			start += tocopy;
+		}
+
+		start &= size - 1;
+		if (rem) {
+			tocopy = min(rem, msc->single_sz - start);
+			if (copy_to_user(buf, msc->base + start, tocopy))
+				return -EFAULT;
+
+			rem -= tocopy;
+		}
+
+		return len - rem;
+	}
+
+	if (copy_to_user(buf, msc->base + start, rem))
+		return -EFAULT;
+
+	return len;
+}
+
+static ssize_t intel_th_msc_read(struct file *file, char __user *buf,
+				 size_t len, loff_t *ppos)
+{
+	struct msc_iter *iter = file->private_data;
+	struct msc *msc = iter->msc;
+	size_t size;
+	loff_t off = *ppos;
+	ssize_t ret = 0;
+
+	if (!atomic_inc_unless_negative(&msc->user_count))
+		return 0;
+
+	if (msc->enabled) {
+		ret = -EBUSY;
+		goto put_count;
+	}
+
+	if (msc->mode == MSC_MODE_SINGLE && !msc->single_wrap)
+		size = msc->single_sz;
+	else
+		size = msc->nr_pages << PAGE_SHIFT;
+
+	if (!size)
+		return 0;
+
+	if (off >= size) {
+		len = 0;
+		goto put_count;
+	}
+	if (off + len >= size)
+		len = size - off;
+
+	if (msc->mode == MSC_MODE_SINGLE) {
+		ret = msc_single_to_user(msc, buf, off, len);
+		if (ret >= 0)
+			*ppos += ret;
+	} else if (msc->mode == MSC_MODE_MULTI) {
+		struct msc_win_to_user_struct u = {
+			.buf	= buf,
+			.offset	= 0,
+		};
+
+		ret = msc_buffer_iterate(iter, len, &u, msc_win_to_user);
+		if (ret >= 0)
+			*ppos = iter->offset;
+	} else {
+		ret = -ENOTSUPP;
+	}
+
+put_count:
+	atomic_dec(&msc->user_count);
+
+	return ret;
+}
+
+/*
+ * vm operations callbacks (vm_ops)
+ */
+
+static void msc_mmap_open(struct vm_area_struct *vma)
+{
+	struct msc_iter *iter = vma->vm_file->private_data;
+	struct msc *msc = iter->msc;
+
+	atomic_inc(&msc->mmap_count);
+}
+
+static void msc_mmap_close(struct vm_area_struct *vma)
+{
+	struct msc_iter *iter = vma->vm_file->private_data;
+	struct msc *msc = iter->msc;
+	unsigned long pg;
+
+	if (!atomic_dec_and_mutex_lock(&msc->mmap_count, &msc->buf_mutex))
+		return;
+
+	/* drop page _counts */
+	for (pg = 0; pg < msc->nr_pages; pg++) {
+		struct page *page = msc_buffer_get_page(msc, pg);
+
+		if (WARN_ON_ONCE(!page))
+			continue;
+
+		if (page->mapping)
+			page->mapping = NULL;
+	}
+
+	/* last mapping -- drop user_count */
+	atomic_dec(&msc->user_count);
+	mutex_unlock(&msc->buf_mutex);
+}
+
+static int msc_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
+{
+	struct msc_iter *iter = vma->vm_file->private_data;
+	struct msc *msc = iter->msc;
+
+	vmf->page = msc_buffer_get_page(msc, vmf->pgoff);
+	if (!vmf->page)
+		return VM_FAULT_SIGBUS;
+
+	get_page(vmf->page);
+	vmf->page->mapping = vma->vm_file->f_mapping;
+	vmf->page->index = vmf->pgoff;
+
+	return 0;
+}
+
+static const struct vm_operations_struct msc_mmap_ops = {
+	.open	= msc_mmap_open,
+	.close	= msc_mmap_close,
+	.fault	= msc_mmap_fault,
+};
+
+static int intel_th_msc_mmap(struct file *file, struct vm_area_struct *vma)
+{
+	unsigned long size = vma->vm_end - vma->vm_start;
+	struct msc_iter *iter = vma->vm_file->private_data;
+	struct msc *msc = iter->msc;
+	int ret = -EINVAL;
+
+	if (!size || offset_in_page(size))
+		return -EINVAL;
+
+	if (vma->vm_pgoff)
+		return -EINVAL;
+
+	/* grab user_count once per mmap; drop in msc_mmap_close() */
+	if (!atomic_inc_unless_negative(&msc->user_count))
+		return -EINVAL;
+
+	if (msc->mode != MSC_MODE_SINGLE &&
+	    msc->mode != MSC_MODE_MULTI)
+		goto out;
+
+	if (size >> PAGE_SHIFT != msc->nr_pages)
+		goto out;
+
+	atomic_set(&msc->mmap_count, 1);
+	ret = 0;
+
+out:
+	if (ret)
+		atomic_dec(&msc->user_count);
+
+	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+	vma->vm_flags |= VM_DONTEXPAND | VM_DONTCOPY;
+	vma->vm_ops = &msc_mmap_ops;
+	return ret;
+}
+
+static const struct file_operations intel_th_msc_fops = {
+	.open		= intel_th_msc_open,
+	.release	= intel_th_msc_release,
+	.read		= intel_th_msc_read,
+	.mmap		= intel_th_msc_mmap,
+	.llseek		= no_llseek,
+};
+
+static int intel_th_msc_init(struct msc *msc)
+{
+	atomic_set(&msc->user_count, -1);
+
+	msc->mode = MSC_MODE_MULTI;
+	mutex_init(&msc->buf_mutex);
+	INIT_LIST_HEAD(&msc->win_list);
+
+	mutex_init(&msc->iter_mutex);
+	INIT_LIST_HEAD(&msc->iter_list);
+
+	msc->burst_len =
+		(ioread32(msc->reg_base + REG_MSU_MSC0CTL) & MSC_LEN) >>
+		__ffs(MSC_LEN);
+
+	return 0;
+}
+
+static const char * const msc_mode[] = {
+	[MSC_MODE_SINGLE]	= "single",
+	[MSC_MODE_MULTI]	= "multi",
+	[MSC_MODE_EXI]		= "ExI",
+	[MSC_MODE_DEBUG]	= "debug",
+};
+
+static ssize_t
+wrap_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct msc *msc = dev_get_drvdata(dev);
+
+	return scnprintf(buf, PAGE_SIZE, "%d\n", msc->wrap);
+}
+
+static ssize_t
+wrap_store(struct device *dev, struct device_attribute *attr, const char *buf,
+	   size_t size)
+{
+	struct msc *msc = dev_get_drvdata(dev);
+	unsigned long val;
+	int ret;
+
+	ret = kstrtoul(buf, 10, &val);
+	if (ret)
+		return ret;
+
+	msc->wrap = !!val;
+
+	return size;
+}
+
+static DEVICE_ATTR_RW(wrap);
+
+static ssize_t
+mode_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct msc *msc = dev_get_drvdata(dev);
+
+	return scnprintf(buf, PAGE_SIZE, "%s\n", msc_mode[msc->mode]);
+}
+
+static ssize_t
+mode_store(struct device *dev, struct device_attribute *attr, const char *buf,
+	   size_t size)
+{
+	struct msc *msc = dev_get_drvdata(dev);
+	size_t len = size;
+	char *cp;
+	int i, ret;
+
+	if (!capable(CAP_SYS_RAWIO))
+		return -EPERM;
+
+	cp = memchr(buf, '\n', len);
+	if (cp)
+		len = cp - buf;
+
+	for (i = 0; i < ARRAY_SIZE(msc_mode); i++)
+		if (!strncmp(msc_mode[i], buf, len))
+			goto found;
+
+	return -EINVAL;
+
+found:
+	mutex_lock(&msc->buf_mutex);
+	ret = msc_buffer_unlocked_free_unless_used(msc);
+	if (!ret)
+		msc->mode = i;
+	mutex_unlock(&msc->buf_mutex);
+
+	return ret ? ret : size;
+}
+
+static DEVICE_ATTR_RW(mode);
+
+static ssize_t
+nr_pages_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct msc *msc = dev_get_drvdata(dev);
+	struct msc_window *win;
+	size_t count = 0;
+
+	mutex_lock(&msc->buf_mutex);
+
+	if (msc->mode == MSC_MODE_SINGLE)
+		count = scnprintf(buf, PAGE_SIZE, "%ld\n", msc->nr_pages);
+	else if (msc->mode == MSC_MODE_MULTI) {
+		list_for_each_entry(win, &msc->win_list, entry) {
+			count += scnprintf(buf + count, PAGE_SIZE - count,
+					   "%d%c", win->nr_blocks,
+					   msc_is_last_win(win) ? '\n' : ',');
+		}
+	} else {
+		count = scnprintf(buf, PAGE_SIZE, "unsupported\n");
+	}
+
+	mutex_unlock(&msc->buf_mutex);
+
+	return count;
+}
+
+static ssize_t
+nr_pages_store(struct device *dev, struct device_attribute *attr,
+	       const char *buf, size_t size)
+{
+	struct msc *msc = dev_get_drvdata(dev);
+	unsigned long val, *win = NULL, *rewin;
+	size_t len = size;
+	const char *p = buf;
+	char *end, *s;
+	int ret, nr_wins = 0;
+
+	if (!capable(CAP_SYS_RAWIO))
+		return -EPERM;
+
+	ret = msc_buffer_free_unless_used(msc);
+	if (ret)
+		return ret;
+
+	/* scan the comma-separated list of allocation sizes */
+	end = memchr(buf, '\n', len);
+	if (end)
+		len = end - buf;
+
+	do {
+		end = memchr(p, ',', len);
+		s = kstrndup(p, end ? end - p : len, GFP_KERNEL);
+		ret = kstrtoul(s, 10, &val);
+		kfree(s);
+
+		if (ret || !val)
+			goto free_win;
+
+		if (nr_wins && msc->mode == MSC_MODE_SINGLE) {
+			ret = -EINVAL;
+			goto free_win;
+		}
+
+		nr_wins++;
+		rewin = krealloc(win, sizeof(*win) * nr_wins, GFP_KERNEL);
+		if (!rewin) {
+			kfree(win);
+			return -ENOMEM;
+		}
+
+		win = rewin;
+		win[nr_wins - 1] = val;
+
+		if (!end)
+			break;
+
+		len -= end - p;
+		p = end + 1;
+	} while (len);
+
+	mutex_lock(&msc->buf_mutex);
+	ret = msc_buffer_alloc(msc, win, nr_wins);
+	mutex_unlock(&msc->buf_mutex);
+
+free_win:
+	kfree(win);
+
+	return ret ? ret : size;
+}
+
+static DEVICE_ATTR_RW(nr_pages);
+
+static struct attribute *msc_output_attrs[] = {
+	&dev_attr_wrap.attr,
+	&dev_attr_mode.attr,
+	&dev_attr_nr_pages.attr,
+	NULL,
+};
+
+static struct attribute_group msc_output_group = {
+	.attrs	= msc_output_attrs,
+};
+
+static int intel_th_msc_probe(struct intel_th_device *thdev)
+{
+	struct device *dev = &thdev->dev;
+	struct resource *res;
+	struct msc *msc;
+	void __iomem *base;
+	int err;
+
+	res = intel_th_device_get_resource(thdev, IORESOURCE_MEM, 0);
+	if (!res)
+		return -ENODEV;
+
+	base = devm_ioremap(dev, res->start, resource_size(res));
+	if (IS_ERR(base))
+		return PTR_ERR(base);
+
+	msc = devm_kzalloc(dev, sizeof(*msc), GFP_KERNEL);
+	if (!msc)
+		return -ENOMEM;
+
+	msc->index = thdev->id;
+
+	msc->thdev = thdev;
+	msc->reg_base = base + msc->index * 0x100;
+
+	err = intel_th_msc_init(msc);
+	if (err)
+		return err;
+
+	err = sysfs_create_group(&dev->kobj, &msc_output_group);
+	if (err)
+		return err;
+
+	dev_set_drvdata(dev, msc);
+
+	return 0;
+}
+
+static void intel_th_msc_remove(struct intel_th_device *thdev)
+{
+	sysfs_remove_group(&thdev->dev.kobj, &msc_output_group);
+}
+
+static struct intel_th_driver intel_th_msc_driver = {
+	.probe	= intel_th_msc_probe,
+	.remove	= intel_th_msc_remove,
+	.activate	= intel_th_msc_activate,
+	.deactivate	= intel_th_msc_deactivate,
+	.fops	= &intel_th_msc_fops,
+	.driver	= {
+		.name	= "msc",
+		.owner	= THIS_MODULE,
+	},
+};
+
+module_driver(intel_th_msc_driver,
+	      intel_th_driver_register,
+	      intel_th_driver_unregister);
+
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION("Intel(R) Trace Hub Memory Storage Unit driver");
+MODULE_AUTHOR("Alexander Shishkin <alexander.shishkin@linux.intel.com>");
diff --git a/drivers/hwtracing/intel_th/msu.h b/drivers/hwtracing/intel_th/msu.h
new file mode 100644
index 0000000000000000000000000000000000000000..9b710e4aa98affe6b1dab3ab3b8cc5b9b3158148
--- /dev/null
+++ b/drivers/hwtracing/intel_th/msu.h
@@ -0,0 +1,116 @@
+/*
+ * Intel(R) Trace Hub Memory Storage Unit (MSU) data structures
+ *
+ * Copyright (C) 2014-2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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 __INTEL_TH_MSU_H__
+#define __INTEL_TH_MSU_H__
+
+enum {
+	REG_MSU_MSUPARAMS	= 0x0000,
+	REG_MSU_MSUSTS		= 0x0008,
+	REG_MSU_MSC0CTL		= 0x0100, /* MSC0 control */
+	REG_MSU_MSC0STS		= 0x0104, /* MSC0 status */
+	REG_MSU_MSC0BAR		= 0x0108, /* MSC0 output base address */
+	REG_MSU_MSC0SIZE	= 0x010c, /* MSC0 output size */
+	REG_MSU_MSC0MWP		= 0x0110, /* MSC0 write pointer */
+	REG_MSU_MSC0NWSA	= 0x011c, /* MSC0 next window start address */
+
+	REG_MSU_MSC1CTL		= 0x0200, /* MSC1 control */
+	REG_MSU_MSC1STS		= 0x0204, /* MSC1 status */
+	REG_MSU_MSC1BAR		= 0x0208, /* MSC1 output base address */
+	REG_MSU_MSC1SIZE	= 0x020c, /* MSC1 output size */
+	REG_MSU_MSC1MWP		= 0x0210, /* MSC1 write pointer */
+	REG_MSU_MSC1NWSA	= 0x021c, /* MSC1 next window start address */
+};
+
+/* MSUSTS bits */
+#define MSUSTS_MSU_INT	BIT(0)
+
+/* MSCnCTL bits */
+#define MSC_EN		BIT(0)
+#define MSC_WRAPEN	BIT(1)
+#define MSC_RD_HDR_OVRD	BIT(2)
+#define MSC_MODE	(BIT(4) | BIT(5))
+#define MSC_LEN		(BIT(8) | BIT(9) | BIT(10))
+
+/* MSC operating modes (MSC_MODE) */
+enum {
+	MSC_MODE_SINGLE	= 0,
+	MSC_MODE_MULTI,
+	MSC_MODE_EXI,
+	MSC_MODE_DEBUG,
+};
+
+/* MSCnSTS bits */
+#define MSCSTS_WRAPSTAT	BIT(1)	/* Wrap occurred */
+#define MSCSTS_PLE	BIT(2)	/* Pipeline Empty */
+
+/*
+ * Multiblock/multiwindow block descriptor
+ */
+struct msc_block_desc {
+	u32	sw_tag;
+	u32	block_sz;
+	u32	next_blk;
+	u32	next_win;
+	u32	res0[4];
+	u32	hw_tag;
+	u32	valid_dw;
+	u32	ts_low;
+	u32	ts_high;
+	u32	res1[4];
+} __packed;
+
+#define MSC_BDESC	sizeof(struct msc_block_desc)
+#define DATA_IN_PAGE	(PAGE_SIZE - MSC_BDESC)
+
+/* MSC multiblock sw tag bits */
+#define MSC_SW_TAG_LASTBLK	BIT(0)
+#define MSC_SW_TAG_LASTWIN	BIT(1)
+
+/* MSC multiblock hw tag bits */
+#define MSC_HW_TAG_TRIGGER	BIT(0)
+#define MSC_HW_TAG_BLOCKWRAP	BIT(1)
+#define MSC_HW_TAG_WINWRAP	BIT(2)
+#define MSC_HW_TAG_ENDBIT	BIT(3)
+
+static inline unsigned long msc_data_sz(struct msc_block_desc *bdesc)
+{
+	if (!bdesc->valid_dw)
+		return 0;
+
+	return bdesc->valid_dw * 4 - MSC_BDESC;
+}
+
+static inline bool msc_block_wrapped(struct msc_block_desc *bdesc)
+{
+	if (bdesc->hw_tag & MSC_HW_TAG_BLOCKWRAP)
+		return true;
+
+	return false;
+}
+
+static inline bool msc_block_last_written(struct msc_block_desc *bdesc)
+{
+	if ((bdesc->hw_tag & MSC_HW_TAG_ENDBIT) ||
+	    (msc_data_sz(bdesc) != DATA_IN_PAGE))
+		return true;
+
+	return false;
+}
+
+/* waiting for Pipeline Empty bit(s) to assert for MSC */
+#define MSC_PLE_WAITLOOP_DEPTH	10000
+
+#endif /* __INTEL_TH_MSU_H__ */