Changelog v3 -> v4:
* https://lore.kernel.org/linux-iio/[email protected]/T/
* added patch 'iio: buffer-dma: reduce the type of block.size to u32'
- resolves error on 64 bit archs; 32 bit block size should be enough
* in patch 'iio: buffer-dma: Add mmap support'
- added 'linux/types.h' include in uapi buffer.h header; an error
shows up when building with 'make allmodconfig'
* in patch 'tools: iio: add example for high-speed buffer support'
- calling ioctl(BOCK_FREE) only if use_high_speed is true
Changelog v2 -> v3:
* https://lore.kernel.org/linux-iio/[email protected]/T/#u
* added 'Documentation: iio: add doc for high-speed buffer API'
* add 'iio: buffer-dma: split iio_dma_buffer_fileio_free() function'
* patch 'iio: buffer-dma: Add mmap support'
- unwind free on error path in iio_dma_buffer_alloc_blocks()
- removed double mm.h include
* patch 'tools: iio: add example for high-speed buffer support'
- call IIO_BUFFER_BLOCK_FREE_IOCTL on the error path of the
enable_high_speed() function
Changelog v1 -> v2:
* https://lore.kernel.org/linux-iio/[email protected]/T/#t
* removed IIO_BUFFER_BLOCK_FLAG_CYCLIC flag; will be added in a later
patch
* removed extra line in tools/iio/iio_generic_buffer.c
* patch 'iio: core: Add mmap interface infrastructure'
added docstrings for new hooks (alloc_blocks, mmap, etc)
This is basically Lars' work adapted from branch:
https://github.com/larsclausen/linux/commits/iio-high-speed-5.10
[hopefully i got the stuff correctly from that branch]
What is different, is that this one is adapted on top of the multibuffer
support (currently at v5) discussed here:
https://lore.kernel.org/linux-iio/[email protected]/T/#t
Also, adapted an example for high-speed/mmap support in
'tools/iio/iio_generic_buffer.c'
The example is adapted from libiio:
https://github.com/analogdevicesinc/libiio/blob/master/local.c#L51
but will all the ioctl()s organized after the one that are reserved
(hopefully) for IIO
Tested that mmap() works.
Moved (artifically) valid buffer0 as buffer2 and the operation still
works.
Alexandru Ardelean (4):
Documentation: iio: add doc for high-speed buffer API
iio: buffer-dma: split iio_dma_buffer_fileio_free() function
iio: buffer-dma: reduce the type of block.size to u32
tools: iio: add example for high-speed buffer support
Lars-Peter Clausen (2):
iio: core: Add mmap interface infrastructure
iio: buffer-dma: Add mmap support
Documentation/iio/iio_high_speed_buffers.rst | 100 ++++++
Documentation/iio/index.rst | 2 +
drivers/iio/buffer/industrialio-buffer-dma.c | 324 ++++++++++++++++--
.../buffer/industrialio-buffer-dmaengine.c | 28 +-
drivers/iio/industrialio-buffer.c | 158 +++++++++
include/linux/iio/buffer-dma.h | 27 +-
include/linux/iio/buffer_impl.h | 23 ++
include/uapi/linux/iio/buffer.h | 51 +++
tools/iio/iio_generic_buffer.c | 185 +++++++++-
9 files changed, 847 insertions(+), 51 deletions(-)
create mode 100644 Documentation/iio/iio_high_speed_buffers.rst
--
2.17.1
From: Lars-Peter Clausen <[email protected]>
Add the necessary infrastructure to the IIO core to support an mmap based
interface to access the capture data.
The advantage of the mmap based interface compared to the read() based
interface is that it avoids an extra copy of the data between kernel and
userspace. This is particular useful for high-speed devices which produce
several megabytes or even gigabytes of data per second.
The data for the mmap interface is managed at the granularity of so called
blocks. A block is a contiguous region of memory (at the moment both
physically and virtually contiguous). Reducing the granularity from byte
level to block level is done to reduce the userspace-kernelspace
synchronization overhead since performing syscalls for each byte at a
data-rate of a few megabytes is not feasible.
This of course leads to a slightly increased latency. For this reason an
application can choose the size of the blocks as well as how many blocks it
allocates. E.g. two blocks would be a traditional double buffering scheme.
But using a higher number might be necessary to avoid underflow/overflow
situations in the presence of scheduling latencies.
A block can either be owned by kernel space or userspace. When owned by
userspace it save to access the data in the block and process it. When
owned by kernel space the block can be in one of 3 states.
It can be in the incoming queue where all blocks submitted from userspace
are placed and are waiting to be processed by the kernel driver.
It can be currently being processed by the kernel driver, this means it is
actively placing capturing data in it (usually using DMA).
Or it can be in the outgoing queue where all blocks that have been
processed by the kernel are placed. Userspace can dequeue the blocks as
necessary.
As part of the interface 5 new IOCTLs to manage the blocks and exchange
them between userspace and kernelspace. The IOCTLs can be accessed through
a open file descriptor to a IIO device.
IIO_BUFFER_BLOCK_ALLOC_IOCTL(struct iio_buffer_block_alloc_req *):
Allocates new blocks. Can be called multiple times if necessary. A newly
allocated block is initially owned by userspace.
IIO_BUFFER_BLOCK_FREE_IOCTL(void):
Frees all previously allocated blocks. If the backing memory of a block is
still in use by a kernel driver (i.e. active DMA transfer) it will be
freed once the kernel driver has released it.
IIO_BUFFER_BLOCK_QUERY_IOCTL(struct iio_buffer_block *):
Queries information about a block. The id of the block about which
information is to be queried needs to be set by userspace.
IIO_BUFFER_BLOCK_ENQUEUE_IOCTL(struct iio_buffer_block *):
Places a block on the incoming queue. This transfers ownership of the
block from userspace to kernelspace. Userspace must populate the id field
of the block to indicate which block to enqueue.
IIO_BUFFER_BLOCK_DEQUEUE_IOCTL(struct iio_buffer_block *):
Removes the first block from the outgoing queue. This transfers ownership
of the block from kernelspace to userspace. Kernelspace will populate all
fields of the block. If the queue is empty and the file descriptor is set
to blocking the IOCTL will block until a new block is available on the
outgoing queue.
To access the data stored in a block by userspace the block must be mapped
to the process's memory. This is done by calling mmap() on the IIO device
file descriptor. Each block has a unique offset assigned to it which should
be passed to the mmap interface. E.g.
mmap(0, block.size, PROT_READ | PROT_WRITE, MAP_SHARED, fd,
block.offset);
A typical workflow for the new interface is:
BLOCK_ALLOC
foreach block
BLOCK_QUERY block
mmap block.data.offset
BLOCK_ENQUEUE block
enable buffer
while !done
BLOCK_DEQUEUE block
process data
BLOCK_ENQUEUE block
disable buffer
BLOCK_FREE
Signed-off-by: Lars-Peter Clausen <[email protected]>
Signed-off-by: Alexandru Ardelean <[email protected]>
---
drivers/iio/industrialio-buffer.c | 158 ++++++++++++++++++++++++++++++
include/linux/iio/buffer-dma.h | 5 -
include/linux/iio/buffer_impl.h | 23 +++++
include/uapi/linux/iio/buffer.h | 46 +++++++++
4 files changed, 227 insertions(+), 5 deletions(-)
diff --git a/drivers/iio/industrialio-buffer.c b/drivers/iio/industrialio-buffer.c
index 4848932d4394..5d641f8adfbd 100644
--- a/drivers/iio/industrialio-buffer.c
+++ b/drivers/iio/industrialio-buffer.c
@@ -17,6 +17,7 @@
#include <linux/fs.h>
#include <linux/cdev.h>
#include <linux/slab.h>
+#include <linux/mm.h>
#include <linux/poll.h>
#include <linux/sched/signal.h>
@@ -1371,6 +1372,12 @@ static void iio_buffer_unregister_legacy_sysfs_groups(struct iio_dev *indio_dev)
kfree(iio_dev_opaque->legacy_scan_el_group.attrs);
}
+static void iio_buffer_free_blocks(struct iio_buffer *buffer)
+{
+ if (buffer->access->free_blocks)
+ buffer->access->free_blocks(buffer);
+}
+
static int iio_buffer_chrdev_release(struct inode *inode, struct file *filep)
{
struct iio_dev_buffer_pair *ib = filep->private_data;
@@ -1381,16 +1388,23 @@ static int iio_buffer_chrdev_release(struct inode *inode, struct file *filep)
kfree(ib);
clear_bit(IIO_BUSY_BIT_POS, &buffer->flags);
+ iio_buffer_free_blocks(buffer);
iio_device_put(indio_dev);
return 0;
}
+static long iio_buffer_ioctl(struct file *filep, unsigned int cmd, unsigned long arg);
+static int iio_buffer_mmap(struct file *filep, struct vm_area_struct *vma);
+
static const struct file_operations iio_buffer_chrdev_fileops = {
.owner = THIS_MODULE,
.llseek = noop_llseek,
.read = iio_buffer_read,
.poll = iio_buffer_poll,
+ .unlocked_ioctl = iio_buffer_ioctl,
+ .compat_ioctl = compat_ptr_ioctl,
+ .mmap = iio_buffer_mmap,
.release = iio_buffer_chrdev_release,
};
@@ -1768,6 +1782,150 @@ void iio_buffer_put(struct iio_buffer *buffer)
}
EXPORT_SYMBOL_GPL(iio_buffer_put);
+static int iio_buffer_query_block(struct iio_buffer *buffer,
+ struct iio_buffer_block __user *user_block)
+{
+ struct iio_buffer_block block;
+ int ret;
+
+ if (!buffer->access->query_block)
+ return -ENOSYS;
+
+ if (copy_from_user(&block, user_block, sizeof(block)))
+ return -EFAULT;
+
+ ret = buffer->access->query_block(buffer, &block);
+ if (ret)
+ return ret;
+
+ if (copy_to_user(user_block, &block, sizeof(block)))
+ return -EFAULT;
+
+ return 0;
+}
+
+static int iio_buffer_dequeue_block(struct iio_dev *indio_dev,
+ struct iio_buffer *buffer,
+ struct iio_buffer_block __user *user_block,
+ bool non_blocking)
+{
+ struct iio_buffer_block block;
+ int ret;
+
+ if (!buffer->access->dequeue_block)
+ return -ENOSYS;
+
+ do {
+ if (!iio_buffer_data_available(buffer)) {
+ if (non_blocking)
+ return -EAGAIN;
+
+ ret = wait_event_interruptible(buffer->pollq,
+ iio_buffer_data_available(buffer) ||
+ indio_dev->info == NULL);
+ if (ret)
+ return ret;
+ if (indio_dev->info == NULL)
+ return -ENODEV;
+ }
+
+ ret = buffer->access->dequeue_block(buffer, &block);
+ if (ret == -EAGAIN && non_blocking)
+ ret = 0;
+ } while (ret);
+
+ if (ret)
+ return ret;
+
+ if (copy_to_user(user_block, &block, sizeof(block)))
+ return -EFAULT;
+
+ return 0;
+}
+
+static int iio_buffer_enqueue_block(struct iio_buffer *buffer,
+ struct iio_buffer_block __user *user_block)
+{
+ struct iio_buffer_block block;
+
+ if (!buffer->access->enqueue_block)
+ return -ENOSYS;
+
+ if (copy_from_user(&block, user_block, sizeof(block)))
+ return -EFAULT;
+
+ return buffer->access->enqueue_block(buffer, &block);
+}
+
+static int iio_buffer_alloc_blocks(struct iio_buffer *buffer,
+ struct iio_buffer_block_alloc_req __user *user_req)
+{
+ struct iio_buffer_block_alloc_req req;
+ int ret;
+
+ if (!buffer->access->alloc_blocks)
+ return -ENOSYS;
+
+ if (copy_from_user(&req, user_req, sizeof(req)))
+ return -EFAULT;
+
+ ret = buffer->access->alloc_blocks(buffer, &req);
+ if (ret)
+ return ret;
+
+ if (copy_to_user(user_req, &req, sizeof(req)))
+ return -EFAULT;
+
+ return 0;
+}
+
+static long iio_buffer_ioctl(struct file *filep, unsigned int cmd, unsigned long arg)
+{
+ bool non_blocking = filep->f_flags & O_NONBLOCK;
+ struct iio_dev_buffer_pair *ib = filep->private_data;
+ struct iio_dev *indio_dev = ib->indio_dev;
+ struct iio_buffer *buffer = ib->buffer;
+
+ if (!buffer || !buffer->access)
+ return -ENODEV;
+
+ switch (cmd) {
+ case IIO_BUFFER_BLOCK_ALLOC_IOCTL:
+ return iio_buffer_alloc_blocks(buffer,
+ (struct iio_buffer_block_alloc_req __user *)arg);
+ case IIO_BUFFER_BLOCK_FREE_IOCTL:
+ iio_buffer_free_blocks(buffer);
+ return 0;
+ case IIO_BUFFER_BLOCK_QUERY_IOCTL:
+ return iio_buffer_query_block(buffer,
+ (struct iio_buffer_block __user *)arg);
+ case IIO_BUFFER_BLOCK_ENQUEUE_IOCTL:
+ return iio_buffer_enqueue_block(buffer,
+ (struct iio_buffer_block __user *)arg);
+ case IIO_BUFFER_BLOCK_DEQUEUE_IOCTL:
+ return iio_buffer_dequeue_block(indio_dev, buffer,
+ (struct iio_buffer_block __user *)arg, non_blocking);
+ }
+ return -EINVAL;
+}
+
+static int iio_buffer_mmap(struct file *filep, struct vm_area_struct *vma)
+{
+ struct iio_dev_buffer_pair *ib = filep->private_data;
+ struct iio_buffer *buffer = ib->buffer;
+
+ if (!buffer->access || !buffer->access->mmap)
+ return -ENODEV;
+
+ if (!(vma->vm_flags & VM_SHARED))
+ return -EINVAL;
+
+ if (!(vma->vm_flags & VM_READ))
+ return -EINVAL;
+
+ return buffer->access->mmap(buffer, vma);
+}
+
/**
* iio_device_attach_buffer - Attach a buffer to a IIO device
* @indio_dev: The device the buffer should be attached to
diff --git a/include/linux/iio/buffer-dma.h b/include/linux/iio/buffer-dma.h
index ff15c61bf319..6564bdcdac66 100644
--- a/include/linux/iio/buffer-dma.h
+++ b/include/linux/iio/buffer-dma.h
@@ -17,11 +17,6 @@ struct iio_dma_buffer_queue;
struct iio_dma_buffer_ops;
struct device;
-struct iio_buffer_block {
- u32 size;
- u32 bytes_used;
-};
-
/**
* enum iio_block_state - State of a struct iio_dma_buffer_block
* @IIO_BLOCK_STATE_DEQUEUED: Block is not queued
diff --git a/include/linux/iio/buffer_impl.h b/include/linux/iio/buffer_impl.h
index 245b32918ae1..1d57dc7ccb4f 100644
--- a/include/linux/iio/buffer_impl.h
+++ b/include/linux/iio/buffer_impl.h
@@ -34,6 +34,18 @@ struct iio_buffer;
* device stops sampling. Calles are balanced with @enable.
* @release: called when the last reference to the buffer is dropped,
* should free all resources allocated by the buffer.
+ * @alloc_blocks: called from userspace via ioctl to allocate blocks
+ * that will be used via the mmap interface.
+ * @free_blocks: called from userspace via ioctl to free all blocks
+ * allocated for this buffer.
+ * @enqueue_block: called from userspace via ioctl to queue this block
+ * to this buffer. Requires a valid block id.
+ * @dequeue_block: called from userspace via ioctl to dequeue this block
+ * from this buffer. Requires a valid block id.
+ * @query_block: called from userspace via ioctl to query the attributes
+ * of this block. Requires a valid block id.
+ * @mmap: mmap hook for this buffer. Userspace mmap() calls will
+ * get routed to this.
* @modes: Supported operating modes by this buffer type
* @flags: A bitmask combination of INDIO_BUFFER_FLAG_*
*
@@ -60,6 +72,17 @@ struct iio_buffer_access_funcs {
void (*release)(struct iio_buffer *buffer);
+ int (*alloc_blocks)(struct iio_buffer *buffer,
+ struct iio_buffer_block_alloc_req *req);
+ int (*free_blocks)(struct iio_buffer *buffer);
+ int (*enqueue_block)(struct iio_buffer *buffer,
+ struct iio_buffer_block *block);
+ int (*dequeue_block)(struct iio_buffer *buffer,
+ struct iio_buffer_block *block);
+ int (*query_block)(struct iio_buffer *buffer,
+ struct iio_buffer_block *block);
+ int (*mmap)(struct iio_buffer *buffer, struct vm_area_struct *vma);
+
unsigned int modes;
unsigned int flags;
};
diff --git a/include/uapi/linux/iio/buffer.h b/include/uapi/linux/iio/buffer.h
index 13939032b3f6..d18692d14aa5 100644
--- a/include/uapi/linux/iio/buffer.h
+++ b/include/uapi/linux/iio/buffer.h
@@ -5,6 +5,52 @@
#ifndef _UAPI_IIO_BUFFER_H_
#define _UAPI_IIO_BUFFER_H_
+#include <linux/types.h>
+
+/**
+ * struct iio_buffer_block_alloc_req - Descriptor for allocating IIO buffer blocks
+ * @type: type of block(s) to allocate (currently unused, reserved)
+ * @size: the size of a single block
+ * @count: the number of blocks to allocate
+ * @id: returned by the request, the number of blocks allocated
+ */
+struct iio_buffer_block_alloc_req {
+ __u32 type;
+ __u32 size;
+ __u32 count;
+ __u32 id;
+};
+
+/* A function will be assigned later for BIT(0) */
+#define IIO_BUFFER_BLOCK_FLAG_RESERVED (1 << 0)
+
+/**
+ * struct iio_buffer_block - Descriptor for a single IIO block
+ * @id: identifier of the block
+ * @size: size of the block
+ * @bytes_used: number of bytes used in this block by a data transfer
+ * @type: type of this block (currently unused, reserved)
+ * @flags: flags for this buffer, set when enqueuing this block
+ * @offset: data offset of this block in a larger memory segment
+ * @timestamp: timestamp for this block
+ */
+struct iio_buffer_block {
+ __u32 id;
+ __u32 size;
+ __u32 bytes_used;
+ __u32 type;
+ __u32 flags;
+ union {
+ __u32 offset;
+ } data;
+ __u64 timestamp;
+};
+
#define IIO_BUFFER_GET_FD_IOCTL _IOWR('i', 0x91, int)
+#define IIO_BUFFER_BLOCK_ALLOC_IOCTL _IOWR('i', 0x92, struct iio_buffer_block_alloc_req)
+#define IIO_BUFFER_BLOCK_FREE_IOCTL _IO('i', 0x93)
+#define IIO_BUFFER_BLOCK_QUERY_IOCTL _IOWR('i', 0x93, struct iio_buffer_block)
+#define IIO_BUFFER_BLOCK_ENQUEUE_IOCTL _IOWR('i', 0x94, struct iio_buffer_block)
+#define IIO_BUFFER_BLOCK_DEQUEUE_IOCTL _IOWR('i', 0x95, struct iio_buffer_block)
#endif /* _UAPI_IIO_BUFFER_H_ */
--
2.17.1
Following a recent update to the IIO buffer infrastructure, this change
adds a basic example on how to access an IIO buffer via the new mmap()
interface.
The ioctl() for the high-speed mode needs to be enabled right from the
start, before setting any parameters via sysfs (length, enable, etc), to
make sure that the mmap mode is used and not the fileio mode.
Signed-off-by: Alexandru Ardelean <[email protected]>
---
tools/iio/iio_generic_buffer.c | 185 +++++++++++++++++++++++++++++++--
1 file changed, 179 insertions(+), 6 deletions(-)
diff --git a/tools/iio/iio_generic_buffer.c b/tools/iio/iio_generic_buffer.c
index 2491c54a5e4f..e7c327fa6091 100644
--- a/tools/iio/iio_generic_buffer.c
+++ b/tools/iio/iio_generic_buffer.c
@@ -31,6 +31,7 @@
#include <stdbool.h>
#include <signal.h>
#include <sys/ioctl.h>
+#include <sys/mman.h>
#include <linux/iio/buffer.h>
#include "iio_utils.h"
@@ -239,6 +240,133 @@ static int enable_disable_all_channels(char *dev_dir_name, int buffer_idx, int e
return 0;
}
+struct mmap_block {
+ struct iio_buffer_block block;
+ void *addr;
+};
+
+static struct mmap_block *enable_high_speed(int buf_fd, unsigned int block_size,
+ int nblocks)
+{
+ struct iio_buffer_block_alloc_req req = { 0 };
+ struct mmap_block *mmaps = NULL;
+ int mmaps_cnt = 0;
+ int i, ret;
+
+ /**
+ * Validate we can do high-speed by issuing BLOCK_FREE ioctl.
+ * If using just BLOCK_ALLOC it's distinguish between ENOSYS
+ * and other error types.
+ */
+ ret = ioctl(buf_fd, IIO_BUFFER_BLOCK_FREE_IOCTL, 0);
+ if (ret < 0) {
+ errno = ENOSYS;
+ return NULL;
+ }
+
+ /* for now, this */
+ req.id = 0;
+ req.type = 0;
+ req.size = block_size;
+ req.count = nblocks;
+
+ ret = ioctl(buf_fd, IIO_BUFFER_BLOCK_ALLOC_IOCTL, &req);
+ if (ret < 0)
+ return NULL;
+
+ if (req.count == 0) {
+ errno = ENOMEM;
+ return NULL;
+ }
+
+ if (req.count < nblocks) {
+ fprintf(stderr, "Requested %d blocks, got %d\n",
+ nblocks, req.count);
+ errno = ENOMEM;
+ return NULL;
+ }
+
+ mmaps = calloc(req.count, sizeof(*mmaps));
+ if (!mmaps) {
+ errno = ENOMEM;
+ return NULL;
+ }
+
+ for (i = 0; i < req.count; i++) {
+ mmaps[i].block.id = i;
+ ret = ioctl(buf_fd, IIO_BUFFER_BLOCK_QUERY_IOCTL, &mmaps[i].block);
+ if (ret < 0)
+ goto error;
+
+ ret = ioctl(buf_fd, IIO_BUFFER_BLOCK_ENQUEUE_IOCTL, &mmaps[i].block);
+ if (ret < 0)
+ goto error;
+
+ mmaps[i].addr = mmap(0, mmaps[i].block.size,
+ PROT_READ | PROT_WRITE, MAP_SHARED,
+ buf_fd, mmaps[i].block.data.offset);
+
+ if (mmaps[i].addr == MAP_FAILED)
+ goto error;
+
+ mmaps_cnt++;
+ }
+
+ return mmaps;
+
+error:
+ for (i = 0; i < mmaps_cnt; i++)
+ munmap(mmaps[i].addr, mmaps[i].block.size);
+ free(mmaps);
+ ioctl(buf_fd, IIO_BUFFER_BLOCK_FREE_IOCTL, 0);
+ return NULL;
+}
+
+static int read_high_speed(int buf_fd, char *data, unsigned int block_size,
+ struct mmap_block *mmaps, unsigned int mmaps_cnt)
+{
+ struct iio_buffer_block block;
+ int ret;
+
+ /**
+ * This is where some buffer-pool management can do wonders,
+ * but for the sake of this sample-code, we're just going to
+ * copy the data and re-enqueue it back
+ */
+ memset(&block, 0, sizeof(block));
+ ret = ioctl(buf_fd, IIO_BUFFER_BLOCK_DEQUEUE_IOCTL, &block);
+ if (ret < 0)
+ return ret;
+
+ /* check for weird conditions */
+ if (block.bytes_used > block_size) {
+ fprintf(stderr,
+ "Got a bigger block (%u) than expected (%u)\n",
+ block.bytes_used, block_size);
+ return -EFBIG;
+ }
+
+ if (block.bytes_used < block_size) {
+ /**
+ * This can be normal, with some real-world data
+ * terminating abruptly. But log it.
+ */
+ fprintf(stderr,
+ "Got a smaller block (%u) than expected (%u)\n",
+ block.bytes_used, block_size);
+ }
+
+ /* memcpy() the data, we lose some more performance here :p */
+ memcpy(data, mmaps[block.id].addr, block.bytes_used);
+
+ /* and re-queue this back */
+ ret = ioctl(buf_fd, IIO_BUFFER_BLOCK_ENQUEUE_IOCTL, &mmaps[block.id].block);
+ if (ret < 0)
+ return ret;
+
+ return block.bytes_used;
+}
+
static void print_usage(void)
{
fprintf(stderr, "Usage: generic_buffer [options]...\n"
@@ -249,6 +377,7 @@ static void print_usage(void)
" -c <n> Do n conversions, or loop forever if n < 0\n"
" -e Disable wait for event (new data)\n"
" -g Use trigger-less mode\n"
+ " -h Use high-speed buffer access\n"
" -l <n> Set buffer length to n samples\n"
" --device-name -n <name>\n"
" --device-num -N <num>\n"
@@ -356,9 +485,15 @@ int main(int argc, char **argv)
struct iio_channel_info *channels = NULL;
+ static bool use_high_speed = false;
+ unsigned int block_size;
+ int nblocks = 16; /* default */
+ int mmaps_cnt = 0;
+ struct mmap_block *mmaps = NULL;
+
register_cleanup();
- while ((c = getopt_long(argc, argv, "aAb:c:egl:n:N:t:T:w:?", longopts,
+ while ((c = getopt_long(argc, argv, "aAb:c:eghl:n:N:t:T:w:?", longopts,
NULL)) != -1) {
switch (c) {
case 'a':
@@ -396,6 +531,9 @@ int main(int argc, char **argv)
case 'g':
notrigger = 1;
break;
+ case 'h':
+ use_high_speed = true;
+ break;
case 'l':
errno = 0;
buf_len = strtoul(optarg, &dummy, 10);
@@ -659,6 +797,29 @@ int main(int argc, char **argv)
goto error;
}
+ scan_size = size_from_channelarray(channels, num_channels);
+ block_size = scan_size * buf_len;
+ /**
+ * Need to enable high-speed before configuring length/enable.
+ * Otherwise, the DMA buffer will work in fileio mode,
+ * and mmap won't work.
+ */
+ if (use_high_speed) {
+ /**
+ * The block_size for one block is the same as 'data', but it
+ * doesn't need to be the same size. It is easier for the sake
+ * of this example.
+ */
+ mmaps = enable_high_speed(buf_fd, block_size, nblocks);
+ if (!mmaps) {
+ fprintf(stderr, "Could not enable high-speed mode\n");
+ ret = -errno;
+ goto error;
+ }
+ mmaps_cnt = nblocks;
+ printf("Using high-speed mode\n");
+ }
+
/* Setup ring buffer parameters */
ret = write_sysfs_int("length", buf_dir_name, buf_len);
if (ret < 0)
@@ -673,8 +834,7 @@ int main(int argc, char **argv)
goto error;
}
- scan_size = size_from_channelarray(channels, num_channels);
- data = malloc(scan_size * buf_len);
+ data = malloc(block_size);
if (!data) {
ret = -ENOMEM;
goto error;
@@ -721,7 +881,13 @@ int main(int argc, char **argv)
toread = 64;
}
- read_size = read(buf_fd, data, toread * scan_size);
+ if (use_high_speed) {
+ read_size = read_high_speed(buf_fd, data, block_size,
+ mmaps, mmaps_cnt);
+ } else {
+ read_size = read(buf_fd, data, toread * scan_size);
+ }
+
if (read_size < 0) {
if (errno == EAGAIN) {
fprintf(stderr, "nothing available\n");
@@ -740,8 +906,15 @@ int main(int argc, char **argv)
if (fd >= 0 && close(fd) == -1)
perror("Failed to close character device");
- if (buf_fd >= 0 && close(buf_fd) == -1)
- perror("Failed to close buffer");
+ for (i = 0; i < mmaps_cnt; i++)
+ munmap(mmaps[i].addr, mmaps[i].block.size);
+ free(mmaps);
+ if (buf_fd >= 0) {
+ if (use_high_speed)
+ ioctl(buf_fd, IIO_BUFFER_BLOCK_FREE_IOCTL, 0);
+ if (close(buf_fd) == -1)
+ perror("Failed to close buffer");
+ }
free(buffer_access);
free(data);
free(buf_dir_name);
--
2.17.1
On Wed, 17 Feb 2021 09:36:32 +0200
Alexandru Ardelean <[email protected]> wrote:
Dropped v3 and applied v4 to the togreg branch of iio.git and pushed
it out as testing for the autobuilders to poke at it and see what
else they can find :)
Jonathan
> Changelog v3 -> v4:
> * https://lore.kernel.org/linux-iio/[email protected]/T/
> * added patch 'iio: buffer-dma: reduce the type of block.size to u32'
> - resolves error on 64 bit archs; 32 bit block size should be enough
> * in patch 'iio: buffer-dma: Add mmap support'
> - added 'linux/types.h' include in uapi buffer.h header; an error
> shows up when building with 'make allmodconfig'
> * in patch 'tools: iio: add example for high-speed buffer support'
> - calling ioctl(BOCK_FREE) only if use_high_speed is true
>
> Changelog v2 -> v3:
> * https://lore.kernel.org/linux-iio/[email protected]/T/#u
> * added 'Documentation: iio: add doc for high-speed buffer API'
> * add 'iio: buffer-dma: split iio_dma_buffer_fileio_free() function'
> * patch 'iio: buffer-dma: Add mmap support'
> - unwind free on error path in iio_dma_buffer_alloc_blocks()
> - removed double mm.h include
> * patch 'tools: iio: add example for high-speed buffer support'
> - call IIO_BUFFER_BLOCK_FREE_IOCTL on the error path of the
> enable_high_speed() function
>
> Changelog v1 -> v2:
> * https://lore.kernel.org/linux-iio/[email protected]/T/#t
> * removed IIO_BUFFER_BLOCK_FLAG_CYCLIC flag; will be added in a later
> patch
> * removed extra line in tools/iio/iio_generic_buffer.c
> * patch 'iio: core: Add mmap interface infrastructure'
> added docstrings for new hooks (alloc_blocks, mmap, etc)
>
> This is basically Lars' work adapted from branch:
> https://github.com/larsclausen/linux/commits/iio-high-speed-5.10
> [hopefully i got the stuff correctly from that branch]
>
> What is different, is that this one is adapted on top of the multibuffer
> support (currently at v5) discussed here:
> https://lore.kernel.org/linux-iio/[email protected]/T/#t
>
> Also, adapted an example for high-speed/mmap support in
> 'tools/iio/iio_generic_buffer.c'
>
> The example is adapted from libiio:
> https://github.com/analogdevicesinc/libiio/blob/master/local.c#L51
> but will all the ioctl()s organized after the one that are reserved
> (hopefully) for IIO
>
> Tested that mmap() works.
> Moved (artifically) valid buffer0 as buffer2 and the operation still
> works.
>
> Alexandru Ardelean (4):
> Documentation: iio: add doc for high-speed buffer API
> iio: buffer-dma: split iio_dma_buffer_fileio_free() function
> iio: buffer-dma: reduce the type of block.size to u32
> tools: iio: add example for high-speed buffer support
>
> Lars-Peter Clausen (2):
> iio: core: Add mmap interface infrastructure
> iio: buffer-dma: Add mmap support
>
> Documentation/iio/iio_high_speed_buffers.rst | 100 ++++++
> Documentation/iio/index.rst | 2 +
> drivers/iio/buffer/industrialio-buffer-dma.c | 324 ++++++++++++++++--
> .../buffer/industrialio-buffer-dmaengine.c | 28 +-
> drivers/iio/industrialio-buffer.c | 158 +++++++++
> include/linux/iio/buffer-dma.h | 27 +-
> include/linux/iio/buffer_impl.h | 23 ++
> include/uapi/linux/iio/buffer.h | 51 +++
> tools/iio/iio_generic_buffer.c | 185 +++++++++-
> 9 files changed, 847 insertions(+), 51 deletions(-)
> create mode 100644 Documentation/iio/iio_high_speed_buffers.rst
>
On Thu, 18 Feb 2021 13:10:45 +0000
Jonathan Cameron <[email protected]> wrote:
> On Wed, 17 Feb 2021 09:36:32 +0200
> Alexandru Ardelean <[email protected]> wrote:
>
> Dropped v3 and applied v4 to the togreg branch of iio.git and pushed
> it out as testing for the autobuilders to poke at it and see what
> else they can find :)
>
> Jonathan
Following an off list discussion, I've dropped this series for now as more
time is needed to consider if the approach is the right way to go.
Jonathan
>
>
> > Changelog v3 -> v4:
> > * https://lore.kernel.org/linux-iio/[email protected]/T/
> > * added patch 'iio: buffer-dma: reduce the type of block.size to u32'
> > - resolves error on 64 bit archs; 32 bit block size should be enough
> > * in patch 'iio: buffer-dma: Add mmap support'
> > - added 'linux/types.h' include in uapi buffer.h header; an error
> > shows up when building with 'make allmodconfig'
> > * in patch 'tools: iio: add example for high-speed buffer support'
> > - calling ioctl(BOCK_FREE) only if use_high_speed is true
> >
> > Changelog v2 -> v3:
> > * https://lore.kernel.org/linux-iio/[email protected]/T/#u
> > * added 'Documentation: iio: add doc for high-speed buffer API'
> > * add 'iio: buffer-dma: split iio_dma_buffer_fileio_free() function'
> > * patch 'iio: buffer-dma: Add mmap support'
> > - unwind free on error path in iio_dma_buffer_alloc_blocks()
> > - removed double mm.h include
> > * patch 'tools: iio: add example for high-speed buffer support'
> > - call IIO_BUFFER_BLOCK_FREE_IOCTL on the error path of the
> > enable_high_speed() function
> >
> > Changelog v1 -> v2:
> > * https://lore.kernel.org/linux-iio/[email protected]/T/#t
> > * removed IIO_BUFFER_BLOCK_FLAG_CYCLIC flag; will be added in a later
> > patch
> > * removed extra line in tools/iio/iio_generic_buffer.c
> > * patch 'iio: core: Add mmap interface infrastructure'
> > added docstrings for new hooks (alloc_blocks, mmap, etc)
> >
> > This is basically Lars' work adapted from branch:
> > https://github.com/larsclausen/linux/commits/iio-high-speed-5.10
> > [hopefully i got the stuff correctly from that branch]
> >
> > What is different, is that this one is adapted on top of the multibuffer
> > support (currently at v5) discussed here:
> > https://lore.kernel.org/linux-iio/[email protected]/T/#t
> >
> > Also, adapted an example for high-speed/mmap support in
> > 'tools/iio/iio_generic_buffer.c'
> >
> > The example is adapted from libiio:
> > https://github.com/analogdevicesinc/libiio/blob/master/local.c#L51
> > but will all the ioctl()s organized after the one that are reserved
> > (hopefully) for IIO
> >
> > Tested that mmap() works.
> > Moved (artifically) valid buffer0 as buffer2 and the operation still
> > works.
> >
> > Alexandru Ardelean (4):
> > Documentation: iio: add doc for high-speed buffer API
> > iio: buffer-dma: split iio_dma_buffer_fileio_free() function
> > iio: buffer-dma: reduce the type of block.size to u32
> > tools: iio: add example for high-speed buffer support
> >
> > Lars-Peter Clausen (2):
> > iio: core: Add mmap interface infrastructure
> > iio: buffer-dma: Add mmap support
> >
> > Documentation/iio/iio_high_speed_buffers.rst | 100 ++++++
> > Documentation/iio/index.rst | 2 +
> > drivers/iio/buffer/industrialio-buffer-dma.c | 324 ++++++++++++++++--
> > .../buffer/industrialio-buffer-dmaengine.c | 28 +-
> > drivers/iio/industrialio-buffer.c | 158 +++++++++
> > include/linux/iio/buffer-dma.h | 27 +-
> > include/linux/iio/buffer_impl.h | 23 ++
> > include/uapi/linux/iio/buffer.h | 51 +++
> > tools/iio/iio_generic_buffer.c | 185 +++++++++-
> > 9 files changed, 847 insertions(+), 51 deletions(-)
> > create mode 100644 Documentation/iio/iio_high_speed_buffers.rst
> >
>