Skip to contentSkip to conversation

userspace MHI client interface driver

PATCH Superseded Hemant Kumar 7 participants 27 messages Nov 28, 2020 – Dec 8, 2020

Subsystems (5)

v13 of 18 Superseded Latest is v18 →

lore
Share & tools
lore
#0 lore raw

v13 cover userspace MHI client interface driver

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

V13:
- Removed LOOPBACK channel from mhi_device_id table from this patch series. Pushing a new patch series to add support for LOOPBACK channel and the user space test application. Also removed the description from kernel documentation.
- Added QMI channel to mhi_device_id table. QMI channel has existing libqmi support from user space.
- Updated kernel Documentation for QMI channel and provided external reference for libqmi.
- Updated device file node name by appending mhi device name only, which already includes mhi controller device name.

V12:
- Added loopback test driver under selftest/drivers/mhi. Updated kernel documentation for the usage of the loopback test application.
- Addressed review comments for renaming variable names, updated inline comments and removed two redundant dev_dbg.

V11:
- Fixed review comments for UCI documentation by expanding TLAs and rewording some sentences.

V10:
- Replaced mutex_lock with mutex_lock_interruptible in read() and write() file ops call back.

V9:
- Renamed dl_lock to dl_pending _lock and pending list to dl_pending for clarity.
- Used read lock to protect cur_buf.
- Change transfer status check logic and only consider 0 and -EOVERFLOW as only success.
- Added __int to module init function.
- Print channel name instead of minor number upon successful probe.

V8:
- Fixed kernel test robot compilation error by changing %lu to %zu for size_t.
- Replaced uci with UCI in Kconfig, commit text, and comments in driver code.
- Fixed minor style related comments.

V7:
- Decoupled uci device and uci channel objects. uci device is associated with device file node. uci channel is associated with MHI channels. uci device refers to uci channel to perform MHI channel operations for device file operations like read() and write(). uci device increments its reference count for every open(). uci device calls mhi_uci_dev_start_chan() to start the MHI channel. uci channel object is tracking number of times MHI channel is referred. This allows to keep the MHI channel in start state until last release() is called. After that uci channel reference count goes to 0 and uci channel clean up is performed which stops the MHI channel. After the last call to release() if driver is removed uci reference count becomes 0 and uci object is cleaned up.
- Use separate uci channel read and write lock to fine grain locking between reader and writer.
- Use uci device lock to synchronize open, release and driver remove.
- Optimize for downlink only or uplink only UCI device.

V6:
- Moved uci.c to mhi directory.
- Updated Kconfig to add module information.
- Updated Makefile to rename uci object file name as mhi_uci
- Removed kref for open count

V5:
- Removed mhi_uci_drv structure.
- Used idr instead of creating global list of uci devices.
- Used kref instead of local ref counting for uci device and open count.
- Removed unlikely macro.

V4:
- Fix locking to protect proper struct members.
- Updated documentation describing uci client driver use cases.
- Fixed uci ref counting in mhi_uci_open for error case.
- Addressed style related review comments.

V3: Added documentation for MHI UCI driver.

V2:
- Added mutex lock to prevent multiple readers to access same
- mhi buffer which can result into use after free.

Hemant Kumar (4):
bus: mhi: core: Add helper API to return number of free TREs
bus: mhi: core: Move MHI_MAX_MTU to external header file
docs: Add documentation for userspace client interface
bus: mhi: Add userspace client interface driver

 Documentation/mhi/index.rst     |   1 +
 Documentation/mhi/uci.rst       |  94 ++++++
 drivers/bus/mhi/Kconfig         |  13 +
 drivers/bus/mhi/Makefile        |   3 +
 drivers/bus/mhi/core/internal.h |   1 -
 drivers/bus/mhi/core/main.c     |  12 +
 drivers/bus/mhi/uci.c           | 665 ++++++++++++++++++++++++++++++++++++++++
 include/linux/mhi.h             |  12 +
 8 files changed, 800 insertions(+), 1 deletion(-)
 create mode 100644 Documentation/mhi/uci.rst
 create mode 100644 drivers/bus/mhi/uci.c
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#1 lore raw

PATCH v13 1/4 bus: mhi: core: Add helper API to return number of free TREs

Introduce mhi_get_free_desc_count() API to return number of TREs available to queue buffer. MHI clients can use this API to know before hand if ring is full without calling queue API.

---
 drivers/bus/mhi/core/main.c | 12 ++++++++++++
 include/linux/mhi.h         |  9 +++++++++
 2 files changed, 21 insertions(+)
drivers/bus/mhi/core/main.c +12 -0
index 4eb93d8..defd579a 100644
@@ -260,6 +260,18 @@ int mhi_destroy_device(struct device *dev, void *data)
 	return 0;
 }
 
+int mhi_get_free_desc_count(struct mhi_device *mhi_dev,
+				enum dma_data_direction dir)
+{
+	struct mhi_controller *mhi_cntrl = mhi_dev->mhi_cntrl;
+	struct mhi_chan *mhi_chan = (dir == DMA_TO_DEVICE) ?
+		mhi_dev->ul_chan : mhi_dev->dl_chan;
+	struct mhi_ring *tre_ring = &mhi_chan->tre_ring;
+
+	return get_nr_avail_ring_elements(mhi_cntrl, tre_ring);
+}
+EXPORT_SYMBOL_GPL(mhi_get_free_desc_count);
+
 void mhi_notify(struct mhi_device *mhi_dev, enum mhi_callback cb_reason)
 {
 	struct mhi_driver *mhi_drv;
include/linux/mhi.h +9 -0
index d31efcf..b3bc966 100644
@@ -597,6 +597,15 @@ void mhi_set_mhi_state(struct mhi_controller *mhi_cntrl,
 void mhi_notify(struct mhi_device *mhi_dev, enum mhi_callback cb_reason);
 
 /**
+ * mhi_get_free_desc_count - Get transfer ring length
+ * Get # of TD available to queue buffers
+ * @mhi_dev: Device associated with the channels
+ * @dir: Direction of the channel
+ */
+int mhi_get_free_desc_count(struct mhi_device *mhi_dev,
+				enum dma_data_direction dir);
+
+/**
  * mhi_prepare_for_power_up - Do pre-initialization before power up.
  *                            This is optional, call this before power up if
  *                            the controller does not want bus framework to
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#2 lore raw

PATCH v13 2/4 bus: mhi: core: Move MHI_MAX_MTU to external header file

In reply to Hemant Kumar

Currently this macro is defined in internal MHI header as a TRE length mask. Moving it to external header allows MHI client drivers to set this upper bound for the transmit buffer size.

---
 drivers/bus/mhi/core/internal.h | 1 -
 include/linux/mhi.h             | 3 +++
 2 files changed, 3 insertions(+), 1 deletion(-)
drivers/bus/mhi/core/internal.h +0 -1
index 6f80ec3..2b9c063 100644
@@ -453,7 +453,6 @@ enum mhi_pm_state {
 #define CMD_EL_PER_RING			128
 #define PRIMARY_CMD_RING		0
 #define MHI_DEV_WAKE_DB			127
-#define MHI_MAX_MTU			0xffff
 #define MHI_RANDOM_U32_NONZERO(bmsk)	(prandom_u32_max(bmsk) + 1)
 
 enum mhi_er_type {
include/linux/mhi.h +3 -0
index b3bc966..fc903b2 100644
@@ -15,6 +15,9 @@
 #include <linux/wait.h>
 #include <linux/workqueue.h>
 
+/* MHI client drivers to set this upper bound for tx buffer */
+#define MHI_MAX_MTU 0xffff
+
 #define MHI_MAX_OEM_PK_HASH_SEGMENTS 16
 
 struct mhi_chan;
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#3 lore raw

PATCH v13 3/4 docs: Add documentation for userspace client interface

In reply to Hemant Kumar

MHI userspace client driver is creating device file node for user application to perform file operations. File operations are handled by MHI core driver. Currently QMI MHI channel is supported by this driver.

---
 Documentation/mhi/index.rst |  1 +
 Documentation/mhi/uci.rst   | 94 +++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 95 insertions(+)
 create mode 100644 Documentation/mhi/uci.rst
Documentation/mhi/index.rst +1 -0
index 1d8dec3..c75a371 100644
@@ -9,6 +9,7 @@ MHI
 
    mhi
    topology
+   uci
 
 .. only::  subproject and html
 
Documentation/mhi/uci.rst new file +94 -0

94 lines not on this page · raw message

Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#4 lore raw

PATCH v13 4/4 bus: mhi: Add userspace client interface driver

In reply to Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

/dev/mhi_<mhi_device_name>

Currently it supports QMI channel.

---
 drivers/bus/mhi/Kconfig  |  13 +
 drivers/bus/mhi/Makefile |   3 +
 drivers/bus/mhi/uci.c    | 665 +++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 681 insertions(+)
 create mode 100644 drivers/bus/mhi/uci.c
drivers/bus/mhi/Kconfig +13 -0
index da5cd0c..5194e8e 100644
@@ -29,3 +29,16 @@ config MHI_BUS_PCI_GENERIC
 	  This driver provides MHI PCI controller driver for devices such as
 	  Qualcomm SDX55 based PCIe modems.
 
+config MHI_UCI
+	tristate "MHI UCI"
+	depends on MHI_BUS
+	help
+	  MHI based Userspace Client Interface (UCI) driver is used for
+	  transferring raw data between host and device using standard file
+	  operations from userspace. Open, read, write, poll and close
+	  operations are supported by this driver. Please check
+	  mhi_uci_match_table for all supported channels that are exposed to
+	  userspace.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called mhi_uci.
drivers/bus/mhi/Makefile +3 -0
index 0a2d778..69f2111 100644
@@ -4,3 +4,6 @@ obj-y += core/
 obj-$(CONFIG_MHI_BUS_PCI_GENERIC) += mhi_pci_generic.o
 mhi_pci_generic-y += pci_generic.o
 
+# MHI client
+mhi_uci-y := uci.o
+obj-$(CONFIG_MHI_UCI) += mhi_uci.o
drivers/bus/mhi/uci.c new file +665 -0

665 lines not on this page · raw message

Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
Reviewed-by on v13 4/4
#5 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

Hi Hemant,

Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

Quoted 4 lines

/dev/mhi_<mhi_device_name>

Currently it supports QMI channel.

Thanks for the update. This patch looks good to me. But as I'm going to apply Loic's "bus: mhi: core: Indexed MHI controller name" patch, you need to update the documentation accordingly.

Quoted 1 line

Thanks,
Mani

Quoted 714 lines
---
 drivers/bus/mhi/Kconfig  |  13 +
 drivers/bus/mhi/Makefile |   3 +
 drivers/bus/mhi/uci.c    | 665 +++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 681 insertions(+)
 create mode 100644 drivers/bus/mhi/uci.c
drivers/bus/mhi/Kconfig +13 -0
index da5cd0c..5194e8e 100644
@@ -29,3 +29,16 @@ config MHI_BUS_PCI_GENERIC
 	  This driver provides MHI PCI controller driver for devices such as
 	  Qualcomm SDX55 based PCIe modems.
 
+config MHI_UCI
+	tristate "MHI UCI"
+	depends on MHI_BUS
+	help
+	  MHI based Userspace Client Interface (UCI) driver is used for
+	  transferring raw data between host and device using standard file
+	  operations from userspace. Open, read, write, poll and close
+	  operations are supported by this driver. Please check
+	  mhi_uci_match_table for all supported channels that are exposed to
+	  userspace.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called mhi_uci.
drivers/bus/mhi/Makefile +3 -0
index 0a2d778..69f2111 100644
@@ -4,3 +4,6 @@ obj-y += core/
 obj-$(CONFIG_MHI_BUS_PCI_GENERIC) += mhi_pci_generic.o
 mhi_pci_generic-y += pci_generic.o
 
+# MHI client
+mhi_uci-y := uci.o
+obj-$(CONFIG_MHI_UCI) += mhi_uci.o
drivers/bus/mhi/uci.c new file +665 -0
new file mode 100644
index 0000000..fb9c183
@@ -0,0 +1,665 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright (c) 2018-2020, The Linux Foundation. All rights reserved.*/
+
+#include <linux/kernel.h>
+#include <linux/mhi.h>
+#include <linux/mod_devicetable.h>
+#include <linux/module.h>
+#include <linux/poll.h>
+
+#define MHI_DEVICE_NAME "mhi"
+#define MHI_UCI_DRIVER_NAME "mhi_uci"
+#define MHI_MAX_UCI_MINORS 128
+
+static DEFINE_IDR(uci_idr);
+static DEFINE_MUTEX(uci_drv_mutex);
+static struct class *uci_dev_class;
+static int uci_dev_major;
+
+/**
+ * struct uci_chan - MHI channel for a UCI device
+ * @udev: associated UCI device object
+ * @ul_wq: wait queue for writer
+ * @write_lock: mutex write lock for ul channel
+ * @dl_wq: wait queue for reader
+ * @read_lock: mutex read lock for dl channel
+ * @dl_pending_lock: spin lock for dl_pending list
+ * @dl_pending: list of dl buffers userspace is waiting to read
+ * @cur_buf: current buffer userspace is reading
+ * @dl_size: size of the current dl buffer userspace is reading
+ * @ref_count: uci_chan reference count
+ */
+struct uci_chan {
+	struct uci_dev *udev;
+	wait_queue_head_t ul_wq;
+
+	/* ul channel lock to synchronize multiple writes */
+	struct mutex write_lock;
+
+	wait_queue_head_t dl_wq;
+
+	/* dl channel lock to synchronize multiple reads */
+	struct mutex read_lock;
+
+	/*
+	 * protects pending list in bh context, channel release, read and
+	 * poll
+	 */
+	spinlock_t dl_pending_lock;
+
+	struct list_head dl_pending;
+	struct uci_buf *cur_buf;
+	size_t dl_size;
+	struct kref ref_count;
+};
+
+/**
+ * struct uci_buf - UCI buffer
+ * @data: data buffer
+ * @len: length of data buffer
+ * @node: list node of the UCI buffer
+ */
+struct uci_buf {
+	void *data;
+	size_t len;
+	struct list_head node;
+};
+
+/**
+ * struct uci_dev - MHI UCI device
+ * @minor: UCI device node minor number
+ * @mhi_dev: associated mhi device object
+ * @uchan: UCI uplink and downlink channel object
+ * @mtu: max TRE buffer length
+ * @enabled: Flag to track the state of the UCI device
+ * @lock: mutex lock to manage uchan object
+ * @ref_count: uci_dev reference count
+ */
+struct uci_dev {
+	unsigned int minor;
+	struct mhi_device *mhi_dev;
+	struct uci_chan *uchan;
+	size_t mtu;
+	bool enabled;
+
+	/* synchronize open, release and driver remove */
+	struct mutex lock;
+	struct kref ref_count;
+};
+
+static void mhi_uci_dev_chan_release(struct kref *ref)
+{
+	struct uci_buf *buf_itr, *tmp;
+	struct uci_chan *uchan =
+		container_of(ref, struct uci_chan, ref_count);
+
+	if (uchan->udev->enabled)
+		mhi_unprepare_from_transfer(uchan->udev->mhi_dev);
+
+	spin_lock_bh(&uchan->dl_pending_lock);
+	list_for_each_entry_safe(buf_itr, tmp, &uchan->dl_pending, node) {
+		list_del(&buf_itr->node);
+		kfree(buf_itr->data);
+	}
+	spin_unlock_bh(&uchan->dl_pending_lock);
+
+	wake_up(&uchan->ul_wq);
+	wake_up(&uchan->dl_wq);
+
+	mutex_lock(&uchan->read_lock);
+	if (uchan->cur_buf)
+		kfree(uchan->cur_buf->data);
+
+	uchan->cur_buf = NULL;
+	mutex_unlock(&uchan->read_lock);
+
+	mutex_destroy(&uchan->write_lock);
+	mutex_destroy(&uchan->read_lock);
+
+	uchan->udev->uchan = NULL;
+	kfree(uchan);
+}
+
+static int mhi_queue_inbound(struct uci_dev *udev)
+{
+	struct mhi_device *mhi_dev = udev->mhi_dev;
+	struct device *dev = &mhi_dev->dev;
+	int nr_desc, i, ret = -EIO;
+	size_t dl_buf_size;
+	void *buf;
+	struct uci_buf *ubuf;
+
+	/*
+	 * skip queuing without error if dl channel is not supported. This
+	 * allows open to succeed for udev, supporting ul only channel.
+	 */
+	if (!udev->mhi_dev->dl_chan)
+		return 0;
+
+	nr_desc = mhi_get_free_desc_count(mhi_dev, DMA_FROM_DEVICE);
+
+	for (i = 0; i < nr_desc; i++) {
+		buf = kmalloc(udev->mtu, GFP_KERNEL);
+		if (!buf)
+			return -ENOMEM;
+
+		dl_buf_size = udev->mtu - sizeof(*ubuf);
+
+		/* save uci_buf info at the end of buf */
+		ubuf = buf + dl_buf_size;
+		ubuf->data = buf;
+
+		dev_dbg(dev, "Allocated buf %d of %d size %zu\n", i, nr_desc,
+			dl_buf_size);
+
+		ret = mhi_queue_buf(mhi_dev, DMA_FROM_DEVICE, buf, dl_buf_size,
+				    MHI_EOT);
+		if (ret) {
+			kfree(buf);
+			dev_err(dev, "Failed to queue buffer %d\n", i);
+			return ret;
+		}
+	}
+
+	return ret;
+}
+
+static int mhi_uci_dev_start_chan(struct uci_dev *udev)
+{
+	int ret = 0;
+	struct uci_chan *uchan;
+
+	mutex_lock(&udev->lock);
+	if (!udev->uchan || !kref_get_unless_zero(&udev->uchan->ref_count)) {
+		uchan = kzalloc(sizeof(*uchan), GFP_KERNEL);
+		if (!uchan) {
+			ret = -ENOMEM;
+			goto error_chan_start;
+		}
+
+		udev->uchan = uchan;
+		uchan->udev = udev;
+		init_waitqueue_head(&uchan->ul_wq);
+		init_waitqueue_head(&uchan->dl_wq);
+		mutex_init(&uchan->write_lock);
+		mutex_init(&uchan->read_lock);
+		spin_lock_init(&uchan->dl_pending_lock);
+		INIT_LIST_HEAD(&uchan->dl_pending);
+
+		ret = mhi_prepare_for_transfer(udev->mhi_dev);
+		if (ret) {
+			dev_err(&udev->mhi_dev->dev, "Error starting transfer channels\n");
+			goto error_chan_cleanup;
+		}
+
+		ret = mhi_queue_inbound(udev);
+		if (ret)
+			goto error_chan_cleanup;
+
+		kref_init(&uchan->ref_count);
+	}
+
+	mutex_unlock(&udev->lock);
+
+	return 0;
+
+error_chan_cleanup:
+	mhi_uci_dev_chan_release(&uchan->ref_count);
+error_chan_start:
+	mutex_unlock(&udev->lock);
+	return ret;
+}
+
+static void mhi_uci_dev_release(struct kref *ref)
+{
+	struct uci_dev *udev =
+		container_of(ref, struct uci_dev, ref_count);
+
+	mutex_destroy(&udev->lock);
+
+	kfree(udev);
+}
+
+static int mhi_uci_open(struct inode *inode, struct file *filp)
+{
+	unsigned int minor = iminor(inode);
+	struct uci_dev *udev = NULL;
+	int ret;
+
+	mutex_lock(&uci_drv_mutex);
+	udev = idr_find(&uci_idr, minor);
+	if (!udev) {
+		pr_debug("uci dev: minor %d not found\n", minor);
+		mutex_unlock(&uci_drv_mutex);
+		return -ENODEV;
+	}
+
+	kref_get(&udev->ref_count);
+	mutex_unlock(&uci_drv_mutex);
+
+	ret = mhi_uci_dev_start_chan(udev);
+	if (ret) {
+		kref_put(&udev->ref_count, mhi_uci_dev_release);
+		return ret;
+	}
+
+	filp->private_data = udev;
+
+	return 0;
+}
+
+static int mhi_uci_release(struct inode *inode, struct file *file)
+{
+	struct uci_dev *udev = file->private_data;
+
+	mutex_lock(&udev->lock);
+	kref_put(&udev->uchan->ref_count, mhi_uci_dev_chan_release);
+	mutex_unlock(&udev->lock);
+
+	kref_put(&udev->ref_count, mhi_uci_dev_release);
+
+	return 0;
+}
+
+static __poll_t mhi_uci_poll(struct file *file, poll_table *wait)
+{
+	struct uci_dev *udev = file->private_data;
+	struct mhi_device *mhi_dev = udev->mhi_dev;
+	struct device *dev = &mhi_dev->dev;
+	struct uci_chan *uchan = udev->uchan;
+	__poll_t mask = 0;
+
+	poll_wait(file, &udev->uchan->ul_wq, wait);
+	poll_wait(file, &udev->uchan->dl_wq, wait);
+
+	if (!udev->enabled) {
+		mask = EPOLLERR;
+		goto done;
+	}
+
+	spin_lock_bh(&uchan->dl_pending_lock);
+	if (!list_empty(&uchan->dl_pending) || uchan->cur_buf)
+		mask |= EPOLLIN | EPOLLRDNORM;
+	spin_unlock_bh(&uchan->dl_pending_lock);
+
+	if (mhi_get_free_desc_count(mhi_dev, DMA_TO_DEVICE) > 0)
+		mask |= EPOLLOUT | EPOLLWRNORM;
+
+	dev_dbg(dev, "Client attempted to poll, returning mask 0x%x\n", mask);
+
+done:
+	return mask;
+}
+
+static ssize_t mhi_uci_write(struct file *file,
+			     const char __user *buf,
+			     size_t count,
+			     loff_t *offp)
+{
+	struct uci_dev *udev = file->private_data;
+	struct mhi_device *mhi_dev = udev->mhi_dev;
+	struct device *dev = &mhi_dev->dev;
+	struct uci_chan *uchan = udev->uchan;
+	size_t bytes_xfered = 0;
+	int ret, nr_desc = 0;
+
+	/* if ul channel is not supported return error */
+	if (!mhi_dev->ul_chan)
+		return -EOPNOTSUPP;
+
+	if (!buf || !count)
+		return -EINVAL;
+
+	dev_dbg(dev, "%s: to xfer: %zu bytes\n", __func__, count);
+
+	if (mutex_lock_interruptible(&uchan->write_lock))
+		return -EINTR;
+
+	while (count) {
+		size_t xfer_size;
+		void *kbuf;
+		enum mhi_flags flags;
+
+		/* wait for free descriptors */
+		ret = wait_event_interruptible(uchan->ul_wq,
+					       (!udev->enabled) ||
+				(nr_desc = mhi_get_free_desc_count(mhi_dev,
+					       DMA_TO_DEVICE)) > 0);
+
+		if (ret == -ERESTARTSYS) {
+			dev_dbg(dev, "Interrupted by a signal in %s, exiting\n",
+				__func__);
+			goto err_mtx_unlock;
+		}
+
+		if (!udev->enabled) {
+			ret = -ENODEV;
+			goto err_mtx_unlock;
+		}
+
+		xfer_size = min_t(size_t, count, udev->mtu);
+		kbuf = kmalloc(xfer_size, GFP_KERNEL);
+		if (!kbuf) {
+			ret = -ENOMEM;
+			goto err_mtx_unlock;
+		}
+
+		ret = copy_from_user(kbuf, buf, xfer_size);
+		if (ret) {
+			kfree(kbuf);
+			ret = -EFAULT;
+			goto err_mtx_unlock;
+		}
+
+		/* if ring is full after this force EOT */
+		if (nr_desc > 1 && (count - xfer_size))
+			flags = MHI_CHAIN;
+		else
+			flags = MHI_EOT;
+
+		ret = mhi_queue_buf(mhi_dev, DMA_TO_DEVICE, kbuf, xfer_size,
+				    flags);
+		if (ret) {
+			kfree(kbuf);
+			goto err_mtx_unlock;
+		}
+
+		bytes_xfered += xfer_size;
+		count -= xfer_size;
+		buf += xfer_size;
+	}
+
+	mutex_unlock(&uchan->write_lock);
+	dev_dbg(dev, "%s: bytes xferred: %zu\n", __func__, bytes_xfered);
+
+	return bytes_xfered;
+
+err_mtx_unlock:
+	mutex_unlock(&uchan->write_lock);
+
+	return ret;
+}
+
+static ssize_t mhi_uci_read(struct file *file,
+			    char __user *buf,
+			    size_t count,
+			    loff_t *ppos)
+{
+	struct uci_dev *udev = file->private_data;
+	struct mhi_device *mhi_dev = udev->mhi_dev;
+	struct uci_chan *uchan = udev->uchan;
+	struct device *dev = &mhi_dev->dev;
+	struct uci_buf *ubuf;
+	size_t rx_buf_size;
+	char *ptr;
+	size_t to_copy;
+	int ret = 0;
+
+	/* if dl channel is not supported return error */
+	if (!mhi_dev->dl_chan)
+		return -EOPNOTSUPP;
+
+	if (!buf)
+		return -EINVAL;
+
+	if (mutex_lock_interruptible(&uchan->read_lock))
+		return -EINTR;
+
+	spin_lock_bh(&uchan->dl_pending_lock);
+	/* No data available to read, wait */
+	if (!uchan->cur_buf && list_empty(&uchan->dl_pending)) {
+		dev_dbg(dev, "No data available to read, waiting\n");
+
+		spin_unlock_bh(&uchan->dl_pending_lock);
+		ret = wait_event_interruptible(uchan->dl_wq,
+					       (!udev->enabled ||
+					      !list_empty(&uchan->dl_pending)));
+
+		if (ret == -ERESTARTSYS) {
+			dev_dbg(dev, "Interrupted by a signal in %s, exiting\n",
+				__func__);
+			goto err_mtx_unlock;
+		}
+
+		if (!udev->enabled) {
+			ret = -ENODEV;
+			goto err_mtx_unlock;
+		}
+		spin_lock_bh(&uchan->dl_pending_lock);
+	}
+
+	/* new read, get the next descriptor from the list */
+	if (!uchan->cur_buf) {
+		ubuf = list_first_entry_or_null(&uchan->dl_pending,
+						struct uci_buf, node);
+		if (!ubuf) {
+			ret = -EIO;
+			goto err_spin_unlock;
+		}
+
+		list_del(&ubuf->node);
+		uchan->cur_buf = ubuf;
+		uchan->dl_size = ubuf->len;
+		dev_dbg(dev, "Got pkt of size: %zu\n", uchan->dl_size);
+	}
+	spin_unlock_bh(&uchan->dl_pending_lock);
+
+	ubuf = uchan->cur_buf;
+
+	/* Copy the buffer to user space */
+	to_copy = min_t(size_t, count, uchan->dl_size);
+	ptr = ubuf->data + (ubuf->len - uchan->dl_size);
+
+	ret = copy_to_user(buf, ptr, to_copy);
+	if (ret) {
+		ret = -EFAULT;
+		goto err_mtx_unlock;
+	}
+
+	dev_dbg(dev, "Copied %zu of %zu bytes\n", to_copy, uchan->dl_size);
+	uchan->dl_size -= to_copy;
+
+	/* we finished with this buffer, queue it back to hardware */
+	if (!uchan->dl_size) {
+		uchan->cur_buf = NULL;
+
+		rx_buf_size = udev->mtu - sizeof(*ubuf);
+		ret = mhi_queue_buf(mhi_dev, DMA_FROM_DEVICE, ubuf->data,
+				    rx_buf_size, MHI_EOT);
+		if (ret) {
+			dev_err(dev, "Failed to recycle element: %d\n", ret);
+			kfree(ubuf->data);
+			goto err_mtx_unlock;
+		}
+	}
+	mutex_unlock(&uchan->read_lock);
+
+	dev_dbg(dev, "%s: Returning %zu bytes\n", __func__, to_copy);
+
+	return to_copy;
+
+err_spin_unlock:
+	spin_unlock_bh(&uchan->dl_pending_lock);
+err_mtx_unlock:
+	mutex_unlock(&uchan->read_lock);
+	return ret;
+}
+
+static const struct file_operations mhidev_fops = {
+	.owner = THIS_MODULE,
+	.open = mhi_uci_open,
+	.release = mhi_uci_release,
+	.read = mhi_uci_read,
+	.write = mhi_uci_write,
+	.poll = mhi_uci_poll,
+};
+
+static void mhi_ul_xfer_cb(struct mhi_device *mhi_dev,
+			   struct mhi_result *mhi_result)
+{
+	struct uci_dev *udev = dev_get_drvdata(&mhi_dev->dev);
+	struct uci_chan *uchan = udev->uchan;
+	struct device *dev = &mhi_dev->dev;
+
+	dev_dbg(dev, "%s: status: %d xfer_len: %zu\n", __func__,
+		mhi_result->transaction_status, mhi_result->bytes_xferd);
+
+	kfree(mhi_result->buf_addr);
+
+	if (!mhi_result->transaction_status)
+		wake_up(&uchan->ul_wq);
+}
+
+static void mhi_dl_xfer_cb(struct mhi_device *mhi_dev,
+			   struct mhi_result *mhi_result)
+{
+	struct uci_dev *udev = dev_get_drvdata(&mhi_dev->dev);
+	struct uci_chan *uchan = udev->uchan;
+	struct device *dev = &mhi_dev->dev;
+	struct uci_buf *ubuf;
+	size_t dl_buf_size = udev->mtu - sizeof(*ubuf);
+
+	dev_dbg(dev, "%s: status: %d receive_len: %zu\n", __func__,
+		mhi_result->transaction_status, mhi_result->bytes_xferd);
+
+	if (mhi_result->transaction_status &&
+	    mhi_result->transaction_status != -EOVERFLOW) {
+		kfree(mhi_result->buf_addr);
+		return;
+	}
+
+	ubuf = mhi_result->buf_addr + dl_buf_size;
+	ubuf->data = mhi_result->buf_addr;
+	ubuf->len = mhi_result->bytes_xferd;
+	spin_lock_bh(&uchan->dl_pending_lock);
+	list_add_tail(&ubuf->node, &uchan->dl_pending);
+	spin_unlock_bh(&uchan->dl_pending_lock);
+
+	wake_up(&uchan->dl_wq);
+}
+
+static int mhi_uci_probe(struct mhi_device *mhi_dev,
+			 const struct mhi_device_id *id)
+{
+	struct uci_dev *udev;
+	struct device *dev;
+	int index;
+
+	udev = kzalloc(sizeof(*udev), GFP_KERNEL);
+	if (!udev)
+		return -ENOMEM;
+
+	kref_init(&udev->ref_count);
+	mutex_init(&udev->lock);
+	udev->mhi_dev = mhi_dev;
+
+	mutex_lock(&uci_drv_mutex);
+	index = idr_alloc(&uci_idr, udev, 0, MHI_MAX_UCI_MINORS, GFP_KERNEL);
+	mutex_unlock(&uci_drv_mutex);
+	if (index < 0) {
+		kfree(udev);
+		return index;
+	}
+
+	udev->minor = index;
+
+	udev->mtu = min_t(size_t, id->driver_data, MHI_MAX_MTU);
+	dev_set_drvdata(&mhi_dev->dev, udev);
+	udev->enabled = true;
+
+	/* create device file node /dev/mhi_<mhi_dev_name> */
+	dev = device_create(uci_dev_class, &mhi_dev->dev,
+			    MKDEV(uci_dev_major, index), udev,
+			    MHI_DEVICE_NAME "_%s", dev_name(&mhi_dev->dev));
+	if (IS_ERR(dev)) {
+		mutex_lock(&uci_drv_mutex);
+		idr_remove(&uci_idr, udev->minor);
+		mutex_unlock(&uci_drv_mutex);
+		dev_set_drvdata(&mhi_dev->dev, NULL);
+		kfree(udev);
+		return PTR_ERR(dev);
+	}
+
+	dev_dbg(&mhi_dev->dev, "probed uci dev: %s\n", id->chan);
+
+	return 0;
+};
+
+static void mhi_uci_remove(struct mhi_device *mhi_dev)
+{
+	struct uci_dev *udev = dev_get_drvdata(&mhi_dev->dev);
+
+	/* disable the node */
+	mutex_lock(&udev->lock);
+	udev->enabled = false;
+
+	/* delete the node to prevent new opens */
+	device_destroy(uci_dev_class, MKDEV(uci_dev_major, udev->minor));
+
+	/* return error for any blocked read or write */
+	if (udev->uchan) {
+		wake_up(&udev->uchan->ul_wq);
+		wake_up(&udev->uchan->dl_wq);
+	}
+	mutex_unlock(&udev->lock);
+
+	mutex_lock(&uci_drv_mutex);
+	idr_remove(&uci_idr, udev->minor);
+	kref_put(&udev->ref_count, mhi_uci_dev_release);
+	mutex_unlock(&uci_drv_mutex);
+}
+
+/* .driver_data stores max mtu */
+static const struct mhi_device_id mhi_uci_match_table[] = {
+	{ .chan = "QMI", .driver_data = 0x1000},
+	{},
+};
+MODULE_DEVICE_TABLE(mhi, mhi_uci_match_table);
+
+static struct mhi_driver mhi_uci_driver = {
+	.id_table = mhi_uci_match_table,
+	.remove = mhi_uci_remove,
+	.probe = mhi_uci_probe,
+	.ul_xfer_cb = mhi_ul_xfer_cb,
+	.dl_xfer_cb = mhi_dl_xfer_cb,
+	.driver = {
+		.name = MHI_UCI_DRIVER_NAME,
+	},
+};
+
+static int __init mhi_uci_init(void)
+{
+	int ret;
+
+	ret = register_chrdev(0, MHI_UCI_DRIVER_NAME, &mhidev_fops);
+	if (ret < 0)
+		return ret;
+
+	uci_dev_major = ret;
+	uci_dev_class = class_create(THIS_MODULE, MHI_UCI_DRIVER_NAME);
+	if (IS_ERR(uci_dev_class)) {
+		unregister_chrdev(uci_dev_major, MHI_UCI_DRIVER_NAME);
+		return PTR_ERR(uci_dev_class);
+	}
+
+	ret = mhi_driver_register(&mhi_uci_driver);
+	if (ret) {
+		class_destroy(uci_dev_class);
+		unregister_chrdev(uci_dev_major, MHI_UCI_DRIVER_NAME);
+	}
+
+	return ret;
+}
+
+static void __exit mhi_uci_exit(void)
+{
+	mhi_driver_unregister(&mhi_uci_driver);
+	class_destroy(uci_dev_class);
+	unregister_chrdev(uci_dev_major, MHI_UCI_DRIVER_NAME);
+	idr_destroy(&uci_idr);
+}
+
+module_init(mhi_uci_init);
+module_exit(mhi_uci_exit);
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION("MHI UCI Driver");
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#6 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

Hi Mani,

Quoted 13 lines · Manivannan Sadhasivam

Hi Hemant,

Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

Quoted 8 lines
Quoted 4 lines

/dev/mhi_<mhi_device_name>

Currently it supports QMI channel.

Thanks for the update. This patch looks good to me. But as I'm going to apply Loic's "bus: mhi: core: Indexed MHI controller name" patch, you need to update the documentation accordingly.

This is what i added in documentation on v13

Device file node is created with format:-

/dev/mhi_<mhi_device_name>

mhi_device_name includes mhi controller name and the name of the MHI channel being used by MHI client in userspace to send or receive data using MHI protocol.

​
Loic's patch is going to update the controller name as indexed controller name, which goes fine with or without his change going first.

For example: With Loic's change name of device node would be /dev/mhi_mhi0_QMI

Without Loic's change it would be

/dev/mhi_0000:00:01.2_QMI

Please let me know if i am missing something.

Thanks,
Hemant

Quoted 8 lines
Quoted 2 lines

Thanks,
Mani

[..]

Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#7 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

In reply to Hemant Kumar
Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

[...]

Quoted 23 lines
drivers/bus/mhi/uci.c new file +23 -0
+struct uci_chan {
+       struct uci_dev *udev;
+       wait_queue_head_t ul_wq;
+
+       /* ul channel lock to synchronize multiple writes */
+       struct mutex write_lock;
+
+       wait_queue_head_t dl_wq;
+
+       /* dl channel lock to synchronize multiple reads */
+       struct mutex read_lock;
+
+       /*
+        * protects pending list in bh context, channel release, read and
+        * poll
+        */
+       spinlock_t dl_pending_lock;
+
+       struct list_head dl_pending;
+       struct uci_buf *cur_buf;
+       size_t dl_size;
+       struct kref ref_count;
+};

[...]

drivers/bus/mhi/uci.c line 81

Quoted 13 lines

+ * struct uci_dev - MHI UCI device
+ * @minor: UCI device node minor number
+ * @mhi_dev: associated mhi device object
+ * @uchan: UCI uplink and downlink channel object
+ * @mtu: max TRE buffer length
+ * @enabled: Flag to track the state of the UCI device
+ * @lock: mutex lock to manage uchan object
+ * @ref_count: uci_dev reference count
+ */

drivers/bus/mhi/uci.c new file +4 -0
+struct uci_dev {
+       unsigned int minor;
+       struct mhi_device *mhi_dev;
+       struct uci_chan *uchan;

Why a pointer to uci_chan and not just plainly integrating the structure here, AFAIU uci_chan describes the channels and is just a subpart of uci_dev. That would reduce the number of dynamic allocations you manage and the extra kref. do you even need a separate structure for this?

[...]

drivers/bus/mhi/uci.c line 173

Quoted 7 lines
drivers/bus/mhi/uci.c new file +7 -0
+static int mhi_uci_dev_start_chan(struct uci_dev *udev)
+{
+       int ret = 0;
+       struct uci_chan *uchan;
+
+       mutex_lock(&udev->lock);
+       if (!udev->uchan || !kref_get_unless_zero(&udev->uchan->ref_count)) {

This test is suspicious, kref_get_unless_zero should be enough to test, right?

if (kref_get_unless_zero(&udev->ref))

    goto skip_init;
Quoted 28 lines
drivers/bus/mhi/uci.c new file +28 -0
+               uchan = kzalloc(sizeof(*uchan), GFP_KERNEL);
+               if (!uchan) {
+                       ret = -ENOMEM;
+                       goto error_chan_start;
+               }
+
+               udev->uchan = uchan;
+               uchan->udev = udev;
+               init_waitqueue_head(&uchan->ul_wq);
+               init_waitqueue_head(&uchan->dl_wq);
+               mutex_init(&uchan->write_lock);
+               mutex_init(&uchan->read_lock);
+               spin_lock_init(&uchan->dl_pending_lock);
+               INIT_LIST_HEAD(&uchan->dl_pending);
+
+               ret = mhi_prepare_for_transfer(udev->mhi_dev);
+               if (ret) {
+                       dev_err(&udev->mhi_dev->dev, "Error starting transfer channels\n");
+                       goto error_chan_cleanup;
+               }
+
+               ret = mhi_queue_inbound(udev);
+               if (ret)
+                       goto error_chan_cleanup;
+
+               kref_init(&uchan->ref_count);
+       }
+

drivers/bus/mhi/uci.c line 211

Quoted 10 lines
drivers/bus/mhi/uci.c new file +10 -0
+       mutex_unlock(&udev->lock);
+
+       return 0;
+
+error_chan_cleanup:
+       mhi_uci_dev_chan_release(&uchan->ref_count);
+error_chan_start:
+       mutex_unlock(&udev->lock);
+       return ret;
+}

Regards,
Loic

#8 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

Hi Loic,

Quoted 54 lines · Loic Poulain
Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

[...]

Quoted 23 lines
drivers/bus/mhi/uci.c new file +23 -0
+struct uci_chan {
+       struct uci_dev *udev;
+       wait_queue_head_t ul_wq;
+
+       /* ul channel lock to synchronize multiple writes */
+       struct mutex write_lock;
+
+       wait_queue_head_t dl_wq;
+
+       /* dl channel lock to synchronize multiple reads */
+       struct mutex read_lock;
+
+       /*
+        * protects pending list in bh context, channel release, read and
+        * poll
+        */
+       spinlock_t dl_pending_lock;
+
+       struct list_head dl_pending;
+       struct uci_buf *cur_buf;
+       size_t dl_size;
+       struct kref ref_count;
+};

[...]

drivers/bus/mhi/uci.c line 81

Quoted 13 lines

+ * struct uci_dev - MHI UCI device
+ * @minor: UCI device node minor number
+ * @mhi_dev: associated mhi device object
+ * @uchan: UCI uplink and downlink channel object
+ * @mtu: max TRE buffer length
+ * @enabled: Flag to track the state of the UCI device
+ * @lock: mutex lock to manage uchan object
+ * @ref_count: uci_dev reference count
+ */

drivers/bus/mhi/uci.c new file +4 -0
+struct uci_dev {
+       unsigned int minor;
+       struct mhi_device *mhi_dev;
+       struct uci_chan *uchan;
Quoted 5 lines

Why a pointer to uci_chan and not just plainly integrating the structure here, AFAIU uci_chan describes the channels and is just a subpart of uci_dev. That would reduce the number of dynamic allocations you manage and the extra kref. do you even need a separate structure for this?

This goes back to one of my patch versions i tried to address concern from Greg. Since we need to ref count the channel as well as the uci device i decoupled the two objects and used two reference counts for two different objects.

Copying from V7 patch history

V7:
- Decoupled uci device and uci channel objects. uci device is associated with device file node. uci channel is associated with MHI channels. uci device refers to uci channel to perform MHI channel operations for device file operations like read() and write(). uci device increments its reference count for every open(). uci device calls mhi_uci_dev_start_chan() to start the MHI channel. uci channel object is tracking number of times MHI channel is referred. This allows to keep the MHI channel in start state until last release() is called. After that uci channel reference count goes to 0 and uci channel clean up is performed which stops the MHI channel. After the last call to release() if driver is removed uci reference count becomes 0 and uci object is cleaned up.

Quoted 12 lines

[...]

drivers/bus/mhi/uci.c line 173

Quoted 7 lines
drivers/bus/mhi/uci.c new file +7 -0
+static int mhi_uci_dev_start_chan(struct uci_dev *udev)
+{
+       int ret = 0;
+       struct uci_chan *uchan;
+
+       mutex_lock(&udev->lock);
+       if (!udev->uchan || !kref_get_unless_zero(&udev->uchan->ref_count)) {

This test is suspicious, kref_get_unless_zero should be enough to test, right?

kref_get_unless_zero is de-referencing uchan->ref_count for the first open uchan is set to NULL, for that NULL check is added for uchan.

Quoted 37 lines

if (kref_get_unless_zero(&udev->ref))

     goto skip_init;
Quoted 33 lines
drivers/bus/mhi/uci.c new file +33 -0
+               uchan = kzalloc(sizeof(*uchan), GFP_KERNEL);
+               if (!uchan) {
+                       ret = -ENOMEM;
+                       goto error_chan_start;
+               }
+
+               udev->uchan = uchan;
+               uchan->udev = udev;
+               init_waitqueue_head(&uchan->ul_wq);
+               init_waitqueue_head(&uchan->dl_wq);
+               mutex_init(&uchan->write_lock);
+               mutex_init(&uchan->read_lock);
+               spin_lock_init(&uchan->dl_pending_lock);
+               INIT_LIST_HEAD(&uchan->dl_pending);
+
+               ret = mhi_prepare_for_transfer(udev->mhi_dev);
+               if (ret) {
+                       dev_err(&udev->mhi_dev->dev, "Error starting transfer channels\n");
+                       goto error_chan_cleanup;
+               }
+
+               ret = mhi_queue_inbound(udev);
+               if (ret)
+                       goto error_chan_cleanup;
+
+               kref_init(&uchan->ref_count);
+       }
+
+       mutex_unlock(&udev->lock);
+
+       return 0;
+
+error_chan_cleanup:
Quoted 9 lines

drivers/bus/mhi/uci.c line 211

Quoted 5 lines
drivers/bus/mhi/uci.c new file +5 -0
+       mhi_uci_dev_chan_release(&uchan->ref_count);
+error_chan_start:
+       mutex_unlock(&udev->lock);
+       return ret;
+}

Regards,
Loic

Thanks,
Hemant

Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
Tested-by
#9 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

Quoted 64 lines · Hemant Kumar

Hi Loic,

Quoted 54 lines · Loic Poulain
Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

[...]

Quoted 23 lines
drivers/bus/mhi/uci.c new file +23 -0
+struct uci_chan {
+       struct uci_dev *udev;
+       wait_queue_head_t ul_wq;
+
+       /* ul channel lock to synchronize multiple writes */
+       struct mutex write_lock;
+
+       wait_queue_head_t dl_wq;
+
+       /* dl channel lock to synchronize multiple reads */
+       struct mutex read_lock;
+
+       /*
+        * protects pending list in bh context, channel release, read and
+        * poll
+        */
+       spinlock_t dl_pending_lock;
+
+       struct list_head dl_pending;
+       struct uci_buf *cur_buf;
+       size_t dl_size;
+       struct kref ref_count;
+};

[...]

drivers/bus/mhi/uci.c line 81

Quoted 13 lines

+ * struct uci_dev - MHI UCI device
+ * @minor: UCI device node minor number
+ * @mhi_dev: associated mhi device object
+ * @uchan: UCI uplink and downlink channel object
+ * @mtu: max TRE buffer length
+ * @enabled: Flag to track the state of the UCI device
+ * @lock: mutex lock to manage uchan object
+ * @ref_count: uci_dev reference count
+ */

drivers/bus/mhi/uci.c new file +4 -0
+struct uci_dev {
+       unsigned int minor;
+       struct mhi_device *mhi_dev;
+       struct uci_chan *uchan;
Quoted 5 lines

Why a pointer to uci_chan and not just plainly integrating the structure here, AFAIU uci_chan describes the channels and is just a subpart of uci_dev. That would reduce the number of dynamic allocations you manage and the extra kref. do you even need a separate structure for this?

Quoted 4 lines

This goes back to one of my patch versions i tried to address concern from Greg. Since we need to ref count the channel as well as the uci device i decoupled the two objects and used two reference counts for two different objects.

What Greg complained about is the two kref in the same structure and that you were using kref as an open() counter. But splitting your struct in two in order to keep the two kref does not make the much code better (and simpler). I'm still a bit puzzled about the driver complexity, it's supposed to be just a passthrough interface to MHI after all.

I would suggest several changes, that IMHO would simplify reviewing:
- Use only one structure representing the 'uci' context (uci_dev)
- Keep the read path simple (mhi_uci_read), do no use an intermediate cur_buf pointer, only dequeue the buffer when it is fully consumed.
- As I commented before, take care of the dl_pending list access concurrency, even in wait_event.
- You don't need to count the number of open() calls, AFAIK, mhi_prepare_for_transfer() simply fails if channels are already started...

For testing purpose, I've implemented those changes on my side (based on your driver):
https://git.linaro.org/landing-teams/working/telit/linux.git/commit/?h=mhi_uci_test&id=45ff60703cc26913061a26260e39cf3ab3e57c2b

Feel free to pick (or not), I'm not going to 'block' that series if others are fine with the current implementation.

Anyway, I've tested your series and it works on my side with libqmi/qmicli (controlling SDX55 modem):

Regards,
Loic

#10 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

Quoted 77 lines · Loic Poulain
Quoted 64 lines · Hemant Kumar

Hi Loic,

Quoted 54 lines · Loic Poulain
Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

[...]

Quoted 23 lines
drivers/bus/mhi/uci.c new file +23 -0
+struct uci_chan {
+       struct uci_dev *udev;
+       wait_queue_head_t ul_wq;
+
+       /* ul channel lock to synchronize multiple writes */
+       struct mutex write_lock;
+
+       wait_queue_head_t dl_wq;
+
+       /* dl channel lock to synchronize multiple reads */
+       struct mutex read_lock;
+
+       /*
+        * protects pending list in bh context, channel release, read and
+        * poll
+        */
+       spinlock_t dl_pending_lock;
+
+       struct list_head dl_pending;
+       struct uci_buf *cur_buf;
+       size_t dl_size;
+       struct kref ref_count;
+};

[...]

drivers/bus/mhi/uci.c line 81

Quoted 13 lines

+ * struct uci_dev - MHI UCI device
+ * @minor: UCI device node minor number
+ * @mhi_dev: associated mhi device object
+ * @uchan: UCI uplink and downlink channel object
+ * @mtu: max TRE buffer length
+ * @enabled: Flag to track the state of the UCI device
+ * @lock: mutex lock to manage uchan object
+ * @ref_count: uci_dev reference count
+ */

drivers/bus/mhi/uci.c new file +4 -0
+struct uci_dev {
+       unsigned int minor;
+       struct mhi_device *mhi_dev;
+       struct uci_chan *uchan;
Quoted 5 lines

Why a pointer to uci_chan and not just plainly integrating the structure here, AFAIU uci_chan describes the channels and is just a subpart of uci_dev. That would reduce the number of dynamic allocations you manage and the extra kref. do you even need a separate structure for this?

Quoted 4 lines

This goes back to one of my patch versions i tried to address concern from Greg. Since we need to ref count the channel as well as the uci device i decoupled the two objects and used two reference counts for two different objects.

What Greg complained about is the two kref in the same structure and that you were using kref as an open() counter. But splitting your struct in two in order to keep the two kref does not make the much code better (and simpler). I'm still a bit puzzled about the driver complexity, it's supposed to be just a passthrough interface to MHI after all.

Quoted 9 lines

I would suggest several changes, that IMHO would simplify reviewing:
- Use only one structure representing the 'uci' context (uci_dev)
- Keep the read path simple (mhi_uci_read), do no use an intermediate cur_buf pointer, only dequeue the buffer when it is fully consumed.
- As I commented before, take care of the dl_pending list access concurrency, even in wait_event.
- You don't need to count the number of open() calls, AFAIK, mhi_prepare_for_transfer() simply fails if channels are already started...

Unless I missed something, you seem to have ignored the root issue that Hemant needs to solve, which is when to call mhi_unprepare_for_transfer(). You can't just call that when close() is called because there might be multiple users, and each one is going to trigger a close(), so you need to know how many close() instances to expect, and only call mhi_unprepare_for_transfer() for the last one.

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
#11 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

Quoted 89 lines · Jeffrey Hugo
Quoted 77 lines · Loic Poulain
Quoted 64 lines · Hemant Kumar

Hi Loic,

Quoted 54 lines · Loic Poulain
Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

[...]

Quoted 23 lines
drivers/bus/mhi/uci.c new file +23 -0
+struct uci_chan {
+       struct uci_dev *udev;
+       wait_queue_head_t ul_wq;
+
+       /* ul channel lock to synchronize multiple writes */
+       struct mutex write_lock;
+
+       wait_queue_head_t dl_wq;
+
+       /* dl channel lock to synchronize multiple reads */
+       struct mutex read_lock;
+
+       /*
+        * protects pending list in bh context, channel release, read and
+        * poll
+        */
+       spinlock_t dl_pending_lock;
+
+       struct list_head dl_pending;
+       struct uci_buf *cur_buf;
+       size_t dl_size;
+       struct kref ref_count;
+};

[...]

drivers/bus/mhi/uci.c line 81

Quoted 13 lines

+ * struct uci_dev - MHI UCI device
+ * @minor: UCI device node minor number
+ * @mhi_dev: associated mhi device object
+ * @uchan: UCI uplink and downlink channel object
+ * @mtu: max TRE buffer length
+ * @enabled: Flag to track the state of the UCI device
+ * @lock: mutex lock to manage uchan object
+ * @ref_count: uci_dev reference count
+ */

drivers/bus/mhi/uci.c new file +4 -0
+struct uci_dev {
+       unsigned int minor;
+       struct mhi_device *mhi_dev;
+       struct uci_chan *uchan;
Quoted 5 lines

Why a pointer to uci_chan and not just plainly integrating the structure here, AFAIU uci_chan describes the channels and is just a subpart of uci_dev. That would reduce the number of dynamic allocations you manage and the extra kref. do you even need a separate structure for this?

Quoted 4 lines

This goes back to one of my patch versions i tried to address concern from Greg. Since we need to ref count the channel as well as the uci device i decoupled the two objects and used two reference counts for two different objects.

What Greg complained about is the two kref in the same structure and that you were using kref as an open() counter. But splitting your struct in two in order to keep the two kref does not make the much code better (and simpler). I'm still a bit puzzled about the driver complexity, it's supposed to be just a passthrough interface to MHI after all.

Quoted 9 lines

I would suggest several changes, that IMHO would simplify reviewing:
- Use only one structure representing the 'uci' context (uci_dev)
- Keep the read path simple (mhi_uci_read), do no use an intermediate cur_buf pointer, only dequeue the buffer when it is fully consumed.
- As I commented before, take care of the dl_pending list access concurrency, even in wait_event.
- You don't need to count the number of open() calls, AFAIK, mhi_prepare_for_transfer() simply fails if channels are already started...

Quoted 6 lines

Unless I missed something, you seem to have ignored the root issue that Hemant needs to solve, which is when to call mhi_unprepare_for_transfer(). You can't just call that when close() is called because there might be multiple users, and each one is going to trigger a close(), so you need to know how many close() instances to expect, and only call mhi_unprepare_for_transfer() for the last one.

That one part of his problem, yes, but if you unconditionally call mhi_prepare_for_transfer in open(), it should fail for subsequent users, and so only one user will successfully open the device.

Regards,
Loic

#12 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

Quoted 97 lines · Loic Poulain
Quoted 89 lines · Jeffrey Hugo
Quoted 77 lines · Loic Poulain
Quoted 64 lines · Hemant Kumar

Hi Loic,

Quoted 54 lines · Loic Poulain
Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

[...]

Quoted 23 lines
drivers/bus/mhi/uci.c new file +23 -0
+struct uci_chan {
+       struct uci_dev *udev;
+       wait_queue_head_t ul_wq;
+
+       /* ul channel lock to synchronize multiple writes */
+       struct mutex write_lock;
+
+       wait_queue_head_t dl_wq;
+
+       /* dl channel lock to synchronize multiple reads */
+       struct mutex read_lock;
+
+       /*
+        * protects pending list in bh context, channel release, read and
+        * poll
+        */
+       spinlock_t dl_pending_lock;
+
+       struct list_head dl_pending;
+       struct uci_buf *cur_buf;
+       size_t dl_size;
+       struct kref ref_count;
+};

[...]

drivers/bus/mhi/uci.c line 81

Quoted 13 lines

+ * struct uci_dev - MHI UCI device
+ * @minor: UCI device node minor number
+ * @mhi_dev: associated mhi device object
+ * @uchan: UCI uplink and downlink channel object
+ * @mtu: max TRE buffer length
+ * @enabled: Flag to track the state of the UCI device
+ * @lock: mutex lock to manage uchan object
+ * @ref_count: uci_dev reference count
+ */

drivers/bus/mhi/uci.c new file +4 -0
+struct uci_dev {
+       unsigned int minor;
+       struct mhi_device *mhi_dev;
+       struct uci_chan *uchan;
Quoted 5 lines

Why a pointer to uci_chan and not just plainly integrating the structure here, AFAIU uci_chan describes the channels and is just a subpart of uci_dev. That would reduce the number of dynamic allocations you manage and the extra kref. do you even need a separate structure for this?

Quoted 4 lines

This goes back to one of my patch versions i tried to address concern from Greg. Since we need to ref count the channel as well as the uci device i decoupled the two objects and used two reference counts for two different objects.

What Greg complained about is the two kref in the same structure and that you were using kref as an open() counter. But splitting your struct in two in order to keep the two kref does not make the much code better (and simpler). I'm still a bit puzzled about the driver complexity, it's supposed to be just a passthrough interface to MHI after all.

Quoted 9 lines

I would suggest several changes, that IMHO would simplify reviewing:
- Use only one structure representing the 'uci' context (uci_dev)
- Keep the read path simple (mhi_uci_read), do no use an intermediate cur_buf pointer, only dequeue the buffer when it is fully consumed.
- As I commented before, take care of the dl_pending list access concurrency, even in wait_event.
- You don't need to count the number of open() calls, AFAIK, mhi_prepare_for_transfer() simply fails if channels are already started...

Quoted 6 lines

Unless I missed something, you seem to have ignored the root issue that Hemant needs to solve, which is when to call mhi_unprepare_for_transfer(). You can't just call that when close() is called because there might be multiple users, and each one is going to trigger a close(), so you need to know how many close() instances to expect, and only call mhi_unprepare_for_transfer() for the last one.

Quoted 3 lines

That one part of his problem, yes, but if you unconditionally call mhi_prepare_for_transfer in open(), it should fail for subsequent users, and so only one user will successfully open the device.

I'm pretty sure that falls under "trying to prevent users from opening a device" which Greg gave a pretty strong NACK to. So, that's not a solution.

So, the complete problem is -

N users need to be able to use the device (and by proxy, the channel) concurrently, but prepare needs to be called on the first user coming into the picture, and unprepare needs to be called on the last user exiting the picture.

Hemant has supported this usecase in every rev of this series I've looked at, but apparently every solution he has proposed to handle this has caused confusion.

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
#13 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

Quoted 99 lines · Jeffrey Hugo
Quoted 97 lines · Loic Poulain
Quoted 95 lines · Jeffrey Hugo
Quoted 77 lines · Loic Poulain
Quoted 64 lines · Hemant Kumar

Hi Loic,

Quoted 54 lines · Loic Poulain
Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

[...]

Quoted 23 lines
(quoted diff) +23 -0
+struct uci_chan {
+       struct uci_dev *udev;
+       wait_queue_head_t ul_wq;
+
+       /* ul channel lock to synchronize multiple writes */
+       struct mutex write_lock;
+
+       wait_queue_head_t dl_wq;
+
+       /* dl channel lock to synchronize multiple reads */
+       struct mutex read_lock;
+
+       /*
+        * protects pending list in bh context, channel release, read and
+        * poll
+        */
+       spinlock_t dl_pending_lock;
+
+       struct list_head dl_pending;
+       struct uci_buf *cur_buf;
+       size_t dl_size;
+       struct kref ref_count;
+};

[...]

Review

Quoted 13 lines

+ * struct uci_dev - MHI UCI device
+ * @minor: UCI device node minor number
+ * @mhi_dev: associated mhi device object
+ * @uchan: UCI uplink and downlink channel object
+ * @mtu: max TRE buffer length
+ * @enabled: Flag to track the state of the UCI device
+ * @lock: mutex lock to manage uchan object
+ * @ref_count: uci_dev reference count
+ */

(quoted diff) +4 -0
+struct uci_dev {
+       unsigned int minor;
+       struct mhi_device *mhi_dev;
+       struct uci_chan *uchan;
Quoted 5 lines

Why a pointer to uci_chan and not just plainly integrating the structure here, AFAIU uci_chan describes the channels and is just a subpart of uci_dev. That would reduce the number of dynamic allocations you manage and the extra kref. do you even need a separate structure for this?

Quoted 4 lines

This goes back to one of my patch versions i tried to address concern from Greg. Since we need to ref count the channel as well as the uci device i decoupled the two objects and used two reference counts for two different objects.

What Greg complained about is the two kref in the same structure and that you were using kref as an open() counter. But splitting your struct in two in order to keep the two kref does not make the much code better (and simpler). I'm still a bit puzzled about the driver complexity, it's supposed to be just a passthrough interface to MHI after all.

Quoted 9 lines

I would suggest several changes, that IMHO would simplify reviewing:
- Use only one structure representing the 'uci' context (uci_dev)
- Keep the read path simple (mhi_uci_read), do no use an intermediate cur_buf pointer, only dequeue the buffer when it is fully consumed.
- As I commented before, take care of the dl_pending list access concurrency, even in wait_event.
- You don't need to count the number of open() calls, AFAIK, mhi_prepare_for_transfer() simply fails if channels are already started...

Unless I missed something, you seem to have ignored the root issue that Hemant needs to solve, which is when to call mhi_unprepare_for_transfer(). You can't just call that when close() is called because there might be multiple users, and each one is going to trigger a close(), so you need to know how many close() instances to expect, and only call mhi_unprepare_for_transfer() for the last one.

Quoted 6 lines
Quoted 3 lines

That one part of his problem, yes, but if you unconditionally call mhi_prepare_for_transfer in open(), it should fail for subsequent users, and so only one user will successfully open the device.

I'm pretty sure that falls under "trying to prevent users from opening a device" which Greg gave a pretty strong NACK to. So, that's not a solution.

That would deserve clarification since other drivers like virtio_console clearly prevent that (guest_connected).

Quoted 11 lines

So, the complete problem is -

N users need to be able to use the device (and by proxy, the channel) concurrently, but prepare needs to be called on the first user coming into the picture, and unprepare needs to be called on the last user exiting the picture.

Hemant has supported this usecase in every rev of this series I've looked at, but apparently every solution he has proposed to handle this has caused confusion.

Understood, but that can be done with a simple counter in that case.

Regards,
Loic

#14 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

Quoted 104 lines · Loic Poulain
Quoted 103 lines · Jeffrey Hugo
Quoted 100 lines · Loic Poulain
Quoted 89 lines · Jeffrey Hugo
Quoted 77 lines · Loic Poulain
Quoted 64 lines · Hemant Kumar

Hi Loic,

Quoted 54 lines · Loic Poulain
Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

[...]

Quoted 23 lines
(quoted diff) +23 -0
+struct uci_chan {
+       struct uci_dev *udev;
+       wait_queue_head_t ul_wq;
+
+       /* ul channel lock to synchronize multiple writes */
+       struct mutex write_lock;
+
+       wait_queue_head_t dl_wq;
+
+       /* dl channel lock to synchronize multiple reads */
+       struct mutex read_lock;
+
+       /*
+        * protects pending list in bh context, channel release, read and
+        * poll
+        */
+       spinlock_t dl_pending_lock;
+
+       struct list_head dl_pending;
+       struct uci_buf *cur_buf;
+       size_t dl_size;
+       struct kref ref_count;
+};

[...]

Review

Quoted 13 lines

+ * struct uci_dev - MHI UCI device
+ * @minor: UCI device node minor number
+ * @mhi_dev: associated mhi device object
+ * @uchan: UCI uplink and downlink channel object
+ * @mtu: max TRE buffer length
+ * @enabled: Flag to track the state of the UCI device
+ * @lock: mutex lock to manage uchan object
+ * @ref_count: uci_dev reference count
+ */

(quoted diff) +4 -0
+struct uci_dev {
+       unsigned int minor;
+       struct mhi_device *mhi_dev;
+       struct uci_chan *uchan;
Quoted 5 lines

Why a pointer to uci_chan and not just plainly integrating the structure here, AFAIU uci_chan describes the channels and is just a subpart of uci_dev. That would reduce the number of dynamic allocations you manage and the extra kref. do you even need a separate structure for this?

Quoted 4 lines

This goes back to one of my patch versions i tried to address concern from Greg. Since we need to ref count the channel as well as the uci device i decoupled the two objects and used two reference counts for two different objects.

What Greg complained about is the two kref in the same structure and that you were using kref as an open() counter. But splitting your struct in two in order to keep the two kref does not make the much code better (and simpler). I'm still a bit puzzled about the driver complexity, it's supposed to be just a passthrough interface to MHI after all.

Quoted 9 lines

I would suggest several changes, that IMHO would simplify reviewing:
- Use only one structure representing the 'uci' context (uci_dev)
- Keep the read path simple (mhi_uci_read), do no use an intermediate cur_buf pointer, only dequeue the buffer when it is fully consumed.
- As I commented before, take care of the dl_pending list access concurrency, even in wait_event.
- You don't need to count the number of open() calls, AFAIK, mhi_prepare_for_transfer() simply fails if channels are already started...

Quoted 6 lines

Unless I missed something, you seem to have ignored the root issue that Hemant needs to solve, which is when to call mhi_unprepare_for_transfer(). You can't just call that when close() is called because there might be multiple users, and each one is going to trigger a close(), so you need to know how many close() instances to expect, and only call mhi_unprepare_for_transfer() for the last one.

That one part of his problem, yes, but if you unconditionally call mhi_prepare_for_transfer in open(), it should fail for subsequent users, and so only one user will successfully open the device.

Quoted 5 lines
Quoted 2 lines

I'm pretty sure that falls under "trying to prevent users from opening a device" which Greg gave a pretty strong NACK to. So, that's not a solution.

That would deserve clarification since other drivers like virtio_console clearly prevent that (guest_connected).

Quoting Greg from the source - https://lkml.org/lkml/2020/9/17/873

"
I told you before, do not try to keep a device node from being opened multiple times, as it will always fail (think about passing file handles around between programs...)

If userspace wants to do this, it will do it. If your driver can't handle that, that's fine, userspace will learn not to do that. But the kernel can not prevent this from happening.
"

Quoted 18 lines
Quoted 4 lines

So, the complete problem is -

Quoted 8 lines

N users need to be able to use the device (and by proxy, the channel) concurrently, but prepare needs to be called on the first user coming into the picture, and unprepare needs to be called on the last user exiting the picture.

Hemant has supported this usecase in every rev of this series I've looked at, but apparently every solution he has proposed to handle this has caused confusion.

Understood, but that can be done with a simple counter in that case.

Regards,
Loic

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
#15 lore raw

Re: [PATCH v13 4/4] bus: mhi: Add userspace client interface driver

Hi Loic,

Quoted 119 lines · Jeffrey Hugo
Quoted 105 lines · Loic Poulain
Quoted 104 lines · Jeffrey Hugo
Quoted 101 lines · Loic Poulain
Quoted 90 lines · Jeffrey Hugo
Quoted 78 lines · Loic Poulain
Quoted 65 lines · Hemant Kumar

Hi Loic,

Quoted 55 lines · Loic Poulain
Quoted 10 lines · Hemant Kumar

This MHI client driver allows userspace clients to transfer raw data between MHI device and host using standard file operations. Driver instantiates UCI device object which is associated to device file node. UCI device object instantiates UCI channel object when device file node is opened. UCI channel object is used to manage MHI channels by calling MHI core APIs for read and write operations. MHI channels are started as part of device open(). MHI channels remain in start state until last release() is called on UCI device file node. Device file node is created with format

[...]

Quoted 24 lines
(quoted diff) +14 -0
+struct uci_chan {
+       struct uci_dev *udev;
+       wait_queue_head_t ul_wq;
+
+       /* ul channel lock to synchronize multiple writes */
+       struct mutex write_lock;
+
+       wait_queue_head_t dl_wq;
+
+       /* dl channel lock to synchronize multiple reads */
+       struct mutex read_lock;
+
+       /*
+        * protects pending list in bh context, channel 

release, read and
+        * poll
+        */
+       spinlock_t dl_pending_lock;

(quoted diff) +6 -0
+
+       struct list_head dl_pending;
+       struct uci_buf *cur_buf;
+       size_t dl_size;
+       struct kref ref_count;
+};

[...]

Quoted 5 lines

+ * struct uci_dev - MHI UCI device
+ * @minor: UCI device node minor number
+ * @mhi_dev: associated mhi device object
+ * @uchan: UCI uplink and downlink channel object
+ * @mtu: max TRE buffer length

Quoted 8 lines

+ * @enabled: Flag to track the state of the UCI device
+ * @lock: mutex lock to manage uchan object
+ * @ref_count: uci_dev reference count
+ */
+struct uci_dev {
+       unsigned int minor;
+       struct mhi_device *mhi_dev;
+       struct uci_chan *uchan;

Quoted 5 lines

Why a pointer to uci_chan and not just plainly integrating the structure here, AFAIU uci_chan describes the channels and is just a subpart of uci_dev. That would reduce the number of dynamic allocations you manage and the extra kref. do you even need a separate structure for this?

Quoted 4 lines

This goes back to one of my patch versions i tried to address concern from Greg. Since we need to ref count the channel as well as the uci device i decoupled the two objects and used two reference counts for two different objects.

What Greg complained about is the two kref in the same structure and that you were using kref as an open() counter. But splitting your struct in two in order to keep the two kref does not make the much code better (and simpler). I'm still a bit puzzled about the driver complexity, it's supposed to be just a passthrough interface to MHI after all.

Quoted 9 lines

I would suggest several changes, that IMHO would simplify reviewing:
- Use only one structure representing the 'uci' context (uci_dev)
- Keep the read path simple (mhi_uci_read), do no use an intermediate cur_buf pointer, only dequeue the buffer when it is fully consumed.
- As I commented before, take care of the dl_pending list access concurrency, even in wait_event.
- You don't need to count the number of open() calls, AFAIK, mhi_prepare_for_transfer() simply fails if channels are already started...

Quoted 6 lines

Unless I missed something, you seem to have ignored the root issue that Hemant needs to solve, which is when to call mhi_unprepare_for_transfer().  You can't just call that when close() is called because there might be multiple users, and each one is going to trigger a close(), so you need to know how many close() instances to expect, and only call mhi_unprepare_for_transfer() for the last one.

That one part of his problem, yes, but if you unconditionally call mhi_prepare_for_transfer in open(), it should fail for subsequent users, and so only one user will successfully open the device.

Quoted 5 lines
Quoted 2 lines

I'm pretty sure that falls under "trying to prevent users from opening a device" which Greg gave a pretty strong NACK to.  So, that's not a solution.

That would deserve clarification since other drivers like virtio_console clearly prevent that (guest_connected).

Quoting Greg from the source - https://lkml.org/lkml/2020/9/17/873

"
I told you before, do not try to keep a device node from being opened multiple times, as it will always fail (think about passing file handles around between programs...)

Quoted 19 lines

If userspace wants to do this, it will do it.  If your driver can't handle that, that's fine, userspace will learn not to do that.  But the kernel can not prevent this from happening.
"

Quoted 14 lines
Quoted 4 lines

So, the complete problem is -

Quoted 8 lines

N users need to be able to use the device (and by proxy, the channel) concurrently, but prepare needs to be called on the first user coming into the picture, and unprepare needs to be called on the last user exiting the picture.

Hemant has supported this usecase in every rev of this series I've looked at, but apparently every solution he has proposed to handle this has caused confusion.

Understood, but that can be done with a simple counter in that case.

If you look at the old original patch set, it was using simple counter and Greg suggested to move away from that and use kref object, and then all incremental changes done using single kref object udev to two separate objects udev and uchan (introduced in v7).

Based on every reviewer's comment i thought driver looked cleaner and moved away from complicated implementations when it was introduced initially. If every one still agrees to that we can move forward with its current form which includes valuable suggestions from all the reviewers.

Thanks for testing my Change on your setup Loic. Mani wants me to update the Documentation patch by modifying the QMI usage example. I am going to upload next revision by updating the documentation.

LOOPBACK test application and its inclusion to mhi device id table would be a separate patch series. Will share the LOOPBACK test application public git hub link in Documentation.

Thanks Jeff for providing the background of the discussions from the past.

Thanks,
Hemant

Quoted 6 lines
Quoted 4 lines

Regards,
Loic

Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
maintainer
#16 lore raw
In reply to Hemant Kumar
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Wait, I thought this was for modems.

Why do WLAN devices need to communicate with user space?

#17 lore raw
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Quoted 4 lines

Wait, I thought this was for modems.

Why do WLAN devices need to communicate with user space?

Why does it matter what type of device it is? Are modems somehow unique in that they are the only type of device that userspace is allowed to interact with?

However, I'll bite. Once such usecase would be QMI. QMI is a generic messaging protocol, and is not strictly limited to the unique operations of a modem.

Another usecase would be Sahara - a custom file transfer protocol used for uploading firmware images, and downloading crashdumps.

Off the top of my head, this driver is useful for modems, wlan, and AI accelerators.

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
maintainer
#18 lore raw
Quoted 9 lines · On Tue, 1 Dec 2020 12:40:50 -0700 Jeffrey Hugo wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Quoted 8 lines
Quoted 4 lines

Wait, I thought this was for modems.

Why do WLAN devices need to communicate with user space?

Why does it matter what type of device it is? Are modems somehow unique in that they are the only type of device that userspace is allowed to interact with?

Yes modems are traditionally highly weird and require some serial device dance I don't even know about.

We have proper interfaces in Linux for configuring WiFi which work across vendors. Having char device access to WiFi would be a step back.

Quoted 6 lines

However, I'll bite. Once such usecase would be QMI. QMI is a generic messaging protocol, and is not strictly limited to the unique operations of a modem.

Another usecase would be Sahara - a custom file transfer protocol used for uploading firmware images, and downloading crashdumps.

Thanks, I was asking for use cases, not which proprietary vendor protocol you can implement over it.

None of the use cases you mention here should require a direct FW - user space backdoor for WLAN.

Quoted 2 lines

Off the top of my head, this driver is useful for modems, wlan, and AI accelerators.

And other Qualcomm products are available as well :/

Kernel is supposed to create abstract interfaces for user space to utilize. I will never understand why kernel is expected to be in business of shipping this sort of vendor backdoors :/

#19 lore raw
Quoted 19 lines · Jakub Kicinski
Quoted 9 lines · On Tue, 1 Dec 2020 12:40:50 -0700 Jeffrey Hugo wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Quoted 8 lines
Quoted 4 lines

Wait, I thought this was for modems.

Why do WLAN devices need to communicate with user space?

Why does it matter what type of device it is? Are modems somehow unique in that they are the only type of device that userspace is allowed to interact with?

Quoted 6 lines

Yes modems are traditionally highly weird and require some serial device dance I don't even know about.

We have proper interfaces in Linux for configuring WiFi which work across vendors. Having char device access to WiFi would be a step back.

So a WLAN device is only ever allowed to do Wi-Fi? It can't also have GPS functionality for example?

Quoted 5 lines
Quoted 5 lines

However, I'll bite. Once such usecase would be QMI. QMI is a generic messaging protocol, and is not strictly limited to the unique operations of a modem.

Quoted 8 lines
Quoted 2 lines

Another usecase would be Sahara - a custom file transfer protocol used for uploading firmware images, and downloading crashdumps.

Thanks, I was asking for use cases, not which proprietary vendor protocol you can implement over it.

None of the use cases you mention here should require a direct FW - user space backdoor for WLAN.

Uploading runtime firmware, with variations based on the runtime mode. Flashing the onboard flash based on cryptographic keys. Accessing configuration data. Accessing device logs. Configuring device logs. Synchronizing the device time reference to Linux local or remote time sources. Enabling debugging/performance hardware. Getting software diagnostic events. Configuring redundancy hardware per workload. Uploading new cryptographic keys. Invalidating cryptographic keys. Uploading factory test data and running factory tests.

Need more?

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
maintainer
#20 lore raw
Quoted 23 lines · On Tue, 1 Dec 2020 13:48:36 -0700 Jeffrey Hugo wrote:
Quoted 22 lines · Jakub Kicinski
Quoted 9 lines · On Tue, 1 Dec 2020 12:40:50 -0700 Jeffrey Hugo wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Quoted 8 lines
Quoted 4 lines

Wait, I thought this was for modems.

Why do WLAN devices need to communicate with user space?

Why does it matter what type of device it is? Are modems somehow unique in that they are the only type of device that userspace is allowed to interact with?

Yes modems are traditionally highly weird and require some serial device dance I don't even know about.

Quoted 6 lines
Quoted 3 lines

We have proper interfaces in Linux for configuring WiFi which work across vendors. Having char device access to WiFi would be a step back.

So a WLAN device is only ever allowed to do Wi-Fi? It can't also have GPS functionality for example?

No, but it's also not true that the only way to implement GPS is by opening a full on command/packet interface between fat proprietary firmware and custom user space (which may or may not be proprietary as well).

Quoted 22 lines
Quoted 4 lines
Quoted 4 lines

However, I'll bite. Once such usecase would be QMI. QMI is a generic messaging protocol, and is not strictly limited to the unique operations of a modem.

Quoted 8 lines
Quoted 2 lines

Another usecase would be Sahara - a custom file transfer protocol used for uploading firmware images, and downloading crashdumps.

Thanks, I was asking for use cases, not which proprietary vendor protocol you can implement over it.

None of the use cases you mention here should require a direct FW - user space backdoor for WLAN.

Uploading runtime firmware, with variations based on the runtime mode. Flashing the onboard flash based on cryptographic keys. Accessing configuration data. Accessing device logs. Configuring device logs. Synchronizing the device time reference to Linux local or remote time sources. Enabling debugging/performance hardware. Getting software diagnostic events. Configuring redundancy hardware per workload. Uploading new cryptographic keys. Invalidating cryptographic keys. Uploading factory test data and running factory tests.

Quoted 1 line

Need more?

This conversation is going nowhere. Are you trying to say that creating a common Linux API for those features is impossible and each vendor should be allowed to add their own proprietary way?

This has been proven incorrect again and again, and Wi-Fi is a good example.

You can do whatever you want for GPS etc. but don't come nowhere near networking with this attitude please.

#21 lore raw
Quoted 28 lines · Jakub Kicinski
Quoted 27 lines · On Tue, 1 Dec 2020 13:48:36 -0700 Jeffrey Hugo wrote:
Quoted 25 lines · Jakub Kicinski
Quoted 9 lines · On Tue, 1 Dec 2020 12:40:50 -0700 Jeffrey Hugo wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Quoted 8 lines
Quoted 4 lines

Wait, I thought this was for modems.

Why do WLAN devices need to communicate with user space?

Why does it matter what type of device it is? Are modems somehow unique in that they are the only type of device that userspace is allowed to interact with?

Yes modems are traditionally highly weird and require some serial device dance I don't even know about.

We have proper interfaces in Linux for configuring WiFi which work across vendors. Having char device access to WiFi would be a step back.

Quoted 7 lines
Quoted 2 lines

So a WLAN device is only ever allowed to do Wi-Fi? It can't also have GPS functionality for example?

No, but it's also not true that the only way to implement GPS is by opening a full on command/packet interface between fat proprietary firmware and custom user space (which may or may not be proprietary as well).

Funny, that exactly what the GPS "API" in the kernel is, although a bit limited to the specifics on the standardized GPS "sentences" and not covering implementation specific configuration.

Quoted 29 lines
Quoted 23 lines
Quoted 5 lines
Quoted 5 lines

However, I'll bite. Once such usecase would be QMI. QMI is a generic messaging protocol, and is not strictly limited to the unique operations of a modem.

Quoted 8 lines
Quoted 2 lines

Another usecase would be Sahara - a custom file transfer protocol used for uploading firmware images, and downloading crashdumps.

Thanks, I was asking for use cases, not which proprietary vendor protocol you can implement over it.

None of the use cases you mention here should require a direct FW - user space backdoor for WLAN.

Uploading runtime firmware, with variations based on the runtime mode. Flashing the onboard flash based on cryptographic keys. Accessing configuration data. Accessing device logs. Configuring device logs. Synchronizing the device time reference to Linux local or remote time sources. Enabling debugging/performance hardware. Getting software diagnostic events. Configuring redundancy hardware per workload. Uploading new cryptographic keys. Invalidating cryptographic keys. Uploading factory test data and running factory tests.

Quoted 1 line

Need more?

This conversation is going nowhere. Are you trying to say that creating a common Linux API for those features is impossible and each vendor should be allowed to add their own proprietary way?

Quoted 6 lines

This has been proven incorrect again and again, and Wi-Fi is a good example.

You can do whatever you want for GPS etc. but don't come nowhere near networking with this attitude please.

No I'm saying (and Bjorn/Mani by the looks of things), that there is commonality in the core features - IP traffic, Wi-Fi, etc but then there are vendor specific things which are either things you don't actually want in the kernel, don't want the kernel doing, or have little commonality between vendors such that attempting to unify them gains you little to nothing.

Over in the networking space, I can see where standardization is plenty useful.

I can't speak for other vendors, but a "modem" or a "wlan" device from Qualcomm is not something that just provides one service. They tend to provide dozens of different functionalities, some of those are "standardized" like wi-fi where common wi-fi interfaces are used. Others are unique to Qualcomm.

The point is "wlan device" is a superset of "wi-fi". You seem to be equating them to be the same in a "shoot first, ask questions later" manner.

This series provides a way for userspace to talk to remote MHI "widgets" for usecases not covered elsewhere. Those "widgets" just happen to commonly provide modem/wlan services, but ones that don't are not excluded.

Regarding not coming near networking, I'd like to remind you it was you that decided to come over here to the non-networking area and try to make this about networking.

Signature · 3 lines
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the
Code Aurora Forum, a Linux Foundation Collaborative Project.
#22 lore raw
Quoted 94 lines · Jeffrey Hugo
Quoted 27 lines · Jakub Kicinski
Quoted 26 lines · On Tue, 1 Dec 2020 13:48:36 -0700 Jeffrey Hugo wrote:
Quoted 24 lines · Jakub Kicinski
Quoted 9 lines · On Tue, 1 Dec 2020 12:40:50 -0700 Jeffrey Hugo wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Quoted 7 lines
Quoted 3 lines

Wait, I thought this was for modems.

Why do WLAN devices need to communicate with user space?

Why does it matter what type of device it is? Are modems somehow unique in that they are the only type of device that userspace is allowed to interact with?

Yes modems are traditionally highly weird and require some serial device dance I don't even know about.

We have proper interfaces in Linux for configuring WiFi which work across vendors. Having char device access to WiFi would be a step back.

Quoted 7 lines
Quoted 2 lines

So a WLAN device is only ever allowed to do Wi-Fi? It can't also have GPS functionality for example?

No, but it's also not true that the only way to implement GPS is by opening a full on command/packet interface between fat proprietary firmware and custom user space (which may or may not be proprietary as well).

Funny, that exactly what the GPS "API" in the kernel is, although a bit limited to the specifics on the standardized GPS "sentences" and not covering implementation specific configuration.

Quoted 29 lines
Quoted 23 lines
Quoted 5 lines
Quoted 5 lines

However, I'll bite. Once such usecase would be QMI. QMI is a generic messaging protocol, and is not strictly limited to the unique operations of a modem.

Quoted 8 lines
Quoted 2 lines

Another usecase would be Sahara - a custom file transfer protocol used for uploading firmware images, and downloading crashdumps.

Thanks, I was asking for use cases, not which proprietary vendor protocol you can implement over it.

None of the use cases you mention here should require a direct FW - user space backdoor for WLAN.

Uploading runtime firmware, with variations based on the runtime mode. Flashing the onboard flash based on cryptographic keys. Accessing configuration data. Accessing device logs. Configuring device logs. Synchronizing the device time reference to Linux local or remote time sources. Enabling debugging/performance hardware. Getting software diagnostic events. Configuring redundancy hardware per workload. Uploading new cryptographic keys. Invalidating cryptographic keys. Uploading factory test data and running factory tests.

Quoted 1 line

Need more?

This conversation is going nowhere. Are you trying to say that creating a common Linux API for those features is impossible and each vendor should be allowed to add their own proprietary way?

Quoted 6 lines

This has been proven incorrect again and again, and Wi-Fi is a good example.

You can do whatever you want for GPS etc. but don't come nowhere near networking with this attitude please.

No I'm saying (and Bjorn/Mani by the looks of things), that there is commonality in the core features - IP traffic, Wi-Fi, etc but then there are vendor specific things which are either things you don't actually want in the kernel, don't want the kernel doing, or have little commonality between vendors such that attempting to unify them gains you little to nothing.

Over in the networking space, I can see where standardization is plenty useful.

I can't speak for other vendors, but a "modem" or a "wlan" device from Qualcomm is not something that just provides one service. They tend to provide dozens of different functionalities, some of those are "standardized" like wi-fi where common wi-fi interfaces are used. Others are unique to Qualcomm.

The point is "wlan device" is a superset of "wi-fi". You seem to be equating them to be the same in a "shoot first, ask questions later" manner.

Quoted 7 lines

This series provides a way for userspace to talk to remote MHI "widgets" for usecases not covered elsewhere. Those "widgets" just happen to commonly provide modem/wlan services, but ones that don't are not excluded.

Regarding not coming near networking, I'd like to remind you it was you that decided to come over here to the non-networking area and try to make this about networking.

Like it or not, but Jakub is absolutely right with his claim that providing user-visible interfaces without any standardization is proven as wrong.

Thanks

Quoted 5 lines

--
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project.

#23 lore raw
Quoted 16 lines · Leon Romanovsky
Quoted 15 lines · Jeffrey Hugo
Quoted 14 lines · Jakub Kicinski
Quoted 13 lines · On Tue, 1 Dec 2020 13:48:36 -0700 Jeffrey Hugo wrote:
Quoted 12 lines · Jakub Kicinski
Quoted 11 lines · On Tue, 1 Dec 2020 12:40:50 -0700 Jeffrey Hugo wrote:
Quoted 10 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Wait, I thought this was for modems.

[...]

Quoted 4 lines

Like it or not, but Jakub is absolutely right with his claim that providing user-visible interfaces without any standardization is proven as wrong.

Everybody agrees with standardizing things but the problem is, the standardization will only happen when more than one person implements the same functionality.

The primary discussion is around the usage of chardev nodes for WLAN but we made it clear that WLAN doesn't need this chardev node for working at all. I agree that the commit message is a bit misleading and I hope that Hemant will fix it in next revision.

Thanks,
Mani

Quoted 7 lines

Thanks

Quoted 5 lines

--
Jeffrey Hugo
Qualcomm Technologies, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project.

#24 lore raw
Quoted 20 lines · Manivannan Sadhasivam
Quoted 16 lines · Leon Romanovsky
Quoted 15 lines · Jeffrey Hugo
Quoted 14 lines · Jakub Kicinski
Quoted 13 lines · On Tue, 1 Dec 2020 13:48:36 -0700 Jeffrey Hugo wrote:
Quoted 12 lines · Jakub Kicinski
Quoted 11 lines · On Tue, 1 Dec 2020 12:40:50 -0700 Jeffrey Hugo wrote:
Quoted 10 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Wait, I thought this was for modems.

[...]

Quoted 8 lines
Quoted 4 lines

Like it or not, but Jakub is absolutely right with his claim that providing user-visible interfaces without any standardization is proven as wrong.

Everybody agrees with standardizing things but the problem is, the standardization will only happen when more than one person implements the same functionality.

From my experience in RDMA and netdev, I can't agree with both of your statements. There are a lot of people who see standardization as a bad thing. Also we are pushing even one person to make user visible interfaces right from the beginning without relation to how wide it will be adopted later.

Thanks

#25 lore raw
In reply to Jakub Kicinski
Quoted 19 lines · Jakub Kicinski
Quoted 9 lines · On Tue, 1 Dec 2020 12:40:50 -0700 Jeffrey Hugo wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Quoted 8 lines
Quoted 4 lines

Wait, I thought this was for modems.

Why do WLAN devices need to communicate with user space?

Why does it matter what type of device it is? Are modems somehow unique in that they are the only type of device that userspace is allowed to interact with?

Quoted 7 lines

Yes modems are traditionally highly weird and require some serial device dance I don't even know about.

We have proper interfaces in Linux for configuring WiFi which work across vendors. Having char device access to WiFi would be a step back.

This is not for configuring the WiFi. This driver is mostly used for modems and the AI accelerator Jeff is working on. But there might be a usecase for WLAN devices as well to collect crash dumps and download fw (typical vendor ways) but having those features are add-ons IMO.

So I think we should not be blocked by those usecases.

Thanks,
Mani

#26 lore raw
In reply to Jakub Kicinski
Quoted 10 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 27 Nov 2020 19:26:02 -0800 Hemant Kumar wrote:

This patch series adds support for UCI driver. UCI driver enables userspace clients to communicate to external MHI devices like modem and WLAN. UCI driver probe creates standard character device file nodes for userspace clients to perform open, read, write, poll and release file operations. These file operations call MHI core layer APIs to perform data transfer using MHI bus to communicate with MHI device. Patch is tested using arm64 based platform.

Wait, I thought this was for modems.

No, this allows exposing particular channels from any type of MHI devices.

For modems there is a legacy control path that uses UCI. But data traffic, (non-legacy) modem control signals and e.g the bearer of GPS data uses in-kernel drivers that are already in place.

Quoted 1 line

Why do WLAN devices need to communicate with user space?

They normally don't, all WLAN operations are dealt with within the kernel. The use case that comes to mind for UCI when it comes to WiFi products is to avoid implementing the Qualcomm debug (diag) protocol in the kernel.

As such I think saying that it can be used to communicate with modem or WLAN devices is misleading. Because while it could be done, it is only used for dealing with optional side-band services on such products - not the actual WiFi and modem functionality.

Regards,
Bjorn