[RFC 0/4] Improving SPI driver latency (vs v3.8.13.14-rt31)

[RFC 0/4] Improving SPI driver latency (vs v3.8.13.14-rt31)

[Jeff Epler]

I recently became interested in realtime access to an SPI device on the
Odroid U3 platform, with a goal of running a repetitive task every 1ms
that performs two SPI transactions. (for http://linuxcnc.org/
interfacing to a http://mesanet.com/ "Anything I/O" card)

Unfortunately, I found that there were frequent large latencies, some
over 10ms, when using /dev/spidev.  This seems to be typical of others'
experience using it (for instance, one can find threads discussing
disappointing RT performance of SPI on the beaglebone and pandaboard; at
least one raspberry pi project chose to implement a pure userspace SPI
driver instead of using spidev)

At all levels of the SPI stack, I found things that could be improved if
lowest delays are the goal.  I doubt that in their current form these
changes are suitable to be incorporated in preempt-rt, but I hope that 
this might spur some discussion that would ultimately lead to better
realtime performance of SPI in the preempt-rt kernel.

As you may know, the kernel's spi driver stack consists of
    spidev    - implementation of the /dev/spidev* device
    spi       - hardware-independent kernel layer
    spi-xyzzy - hardware-dependent driver for xyzzy hardware
                (s3c64xx in my device)

I fixed causes of latency at each layer
 * First, I eliminated per-ioctl memory allocations in spidev 
 * Second, I made __spi_sync *actually* synchronous, rather than
   being a wrapper over spi_async + wait_for_completion; and changed
   spidev to use spi_sync
 * Third, in the hardware-dependent code I moved DMA acquisition to
   device initialization time rather than transaction time

I did not quite achieve my goal of a 1ms repetitive rate yet, but with
these changes I have run for 12+ hours at a rate of 3 transactions per
2ms with acceptable worst-case performance---under 250us for the biggest
transaction, and 465us for all three (they have different sizes), with
typical figures of more like 200us for all three transactions.  This is
in contrast to the original performance, in which transactions taking
over 10ms were seen multiple times per hour.  (12 hours is about 64
million spi transations)

(I changed from talking about 2 transactions to 3, because for an
unrelated reason the communication in my program is currently divided
into 3 SPI transactions when two would do)

I know that 3.8 is by no means current, but 3.8.y is the default kernel
shipped by hardkernel for their U3 devices so it was a rational version
choice for me.  I did skim spi changes from version 3.8 to the present
and didn't see anything that looked like it was directed at improving
SPI latency, though the underlying code has probably changed enough over
time that I assume my patches wouldn't actually apply at the tip of the
latest and greatest branches.

PS The fact that the first PREEMPT-RT kernel I built for the odroid
worked and has basically good latency (until trying to talk to the
hardware :-/) impressed the heck out of me.

Jeff Epler (4):
  spi: reenable sync SPI transfers
  spidev: Avoid runtime memory allocations
  spidev: actually use synchronous transfers
  spidev-s3c64xx: allocate dma channel at startup

 drivers/spi/spi-s3c64xx.c | 15 +++++++--------
 drivers/spi/spi.c         | 22 ++++++++++++++++------
 drivers/spi/spidev.c      | 43 ++++++++++++++++++++++++++++++-------------
 3 files changed, 53 insertions(+), 27 deletions(-)

-- 
2.0.1

[PATCH 1/4] spi: reenable sync SPI transfers

[Jeff Epler]

one source of excess latency in hm2_spi / hal_spidev is latency due
to async transfers.  Make the __spi_sync primitive actually synchronous,
rather than building on an asynchronous primitive.
---
 drivers/spi/spi.c | 22 ++++++++++++++++------
 1 file changed, 16 insertions(+), 6 deletions(-)

diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c
index 19ee901..55404b5 100644
--- a/drivers/spi/spi.c
+++ b/drivers/spi/spi.c
@@ -664,10 +664,18 @@ void spi_finalize_current_message(struct spi_master *master)
 	struct spi_message *mesg;
 	unsigned long flags;
 
+        if(!master->cur_msg)
+                return;
+
 	spin_lock_irqsave(&master->queue_lock, flags);
 	mesg = master->cur_msg;
 	master->cur_msg = NULL;
 
+        if(!mesg) {
+                spin_unlock_irqrestore(&master->queue_lock, flags);
+                return;
+        }
+
 	queue_kthread_work(&master->kworker, &master->pump_messages);
 	spin_unlock_irqrestore(&master->queue_lock, flags);
 
@@ -1490,24 +1498,26 @@ static int __spi_sync(struct spi_device *spi, struct spi_message *message,
 		      int bus_locked)
 {
 	DECLARE_COMPLETION_ONSTACK(done);
-	int status;
+	int status = 0;
 	struct spi_master *master = spi->master;
 
 	message->complete = spi_complete;
 	message->context = &done;
+	message->spi = spi;
 
 	if (!bus_locked)
 		mutex_lock(&master->bus_lock_mutex);
 
-	status = spi_async_locked(spi, message);
+        if(master->prepare_transfer_hardware)
+            status = master->prepare_transfer_hardware(master);
+        if(status >= 0)
+            status = master->transfer_one_message(master, message);
+        if(status >= 0 && master->unprepare_transfer_hardware)
+            status = master->unprepare_transfer_hardware(master);
 
 	if (!bus_locked)
 		mutex_unlock(&master->bus_lock_mutex);
 
-	if (status == 0) {
-		wait_for_completion(&done);
-		status = message->status;
-	}
 	message->context = NULL;
 	return status;
 }
-- 
2.0.1

[PATCH 2/4] spidev: Avoid runtime memory allocations

[Jeff Epler]

allocating bounce and k_xfers at every iteration was a source of
excess delays.
---
 drivers/spi/spidev.c | 35 +++++++++++++++++++++++++++++------
 1 file changed, 29 insertions(+), 6 deletions(-)

diff --git a/drivers/spi/spidev.c b/drivers/spi/spidev.c
index 2e0655d..dda7632 100644
--- a/drivers/spi/spidev.c
+++ b/drivers/spi/spidev.c
@@ -85,6 +85,8 @@ struct spidev_data {
 	struct mutex		buf_lock;
 	unsigned		users;
 	u8			*buffer;
+        struct spi_ioc_transfer *bounce;
+        struct spi_transfer     *k_xfers;
 };
 
 static LIST_HEAD(device_list);
@@ -227,9 +229,9 @@ static int spidev_message(struct spidev_data *spidev,
 	int			status = -EFAULT;
 
 	spi_message_init(&msg);
-	k_xfers = kcalloc(n_xfers, sizeof(*k_tmp), GFP_KERNEL);
-	if (k_xfers == NULL)
-		return -ENOMEM;
+        if (n_xfers * sizeof(*k_tmp) > bufsiz)
+                return -ENOMEM;
+	k_xfers = spidev->k_xfers;
 
 	/* Construct spi_message, copying any tx data to bounce buffer.
 	 * We walk the array of user-provided transfers, using each one
@@ -302,7 +304,6 @@ static int spidev_message(struct spidev_data *spidev,
 	status = total;
 
 done:
-	kfree(k_xfers);
 	return status;
 }
 
@@ -451,8 +452,12 @@ spidev_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 		if (n_ioc == 0)
 			break;
 
+                if(tmp > bufsiz) {
+			retval = -EINVAL;
+			break;
+                }
 		/* copy into scratch area */
-		ioc = kmalloc(tmp, GFP_KERNEL);
+		ioc = spidev->bounce;
 		if (!ioc) {
 			retval = -ENOMEM;
 			break;
@@ -465,7 +470,6 @@ spidev_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 
 		/* translate to spi_message, execute */
 		retval = spidev_message(spidev, ioc, n_ioc);
-		kfree(ioc);
 		break;
 	}
 
@@ -504,6 +508,19 @@ static int spidev_open(struct inode *inode, struct file *filp)
 				dev_dbg(&spidev->spi->dev, "open/ENOMEM\n");
 				status = -ENOMEM;
 			}
+			spidev->bounce = kmalloc(bufsiz, GFP_KERNEL);
+			if (!spidev->bounce) {
+				dev_dbg(&spidev->spi->dev, "open/ENOMEM\n");
+                                kfree(spidev->buffer);
+				status = -ENOMEM;
+			}
+			spidev->k_xfers = kmalloc(bufsiz, GFP_KERNEL);
+			if (!spidev->k_xfers) {
+				dev_dbg(&spidev->spi->dev, "open/ENOMEM\n");
+                                kfree(spidev->buffer);
+                                kfree(spidev->bounce);
+				status = -ENOMEM;
+			}
 		}
 		if (status == 0) {
 			spidev->users++;
@@ -534,6 +551,12 @@ static int spidev_release(struct inode *inode, struct file *filp)
 		kfree(spidev->buffer);
 		spidev->buffer = NULL;
 
+		kfree(spidev->bounce);
+		spidev->bounce = NULL;
+
+		kfree(spidev->k_xfers);
+		spidev->k_xfers = NULL;
+
 		/* ... after we unbound from the underlying device? */
 		spin_lock_irq(&spidev->spi_lock);
 		dofree = (spidev->spi == NULL);
-- 
2.0.1

[PATCH 3/4] spidev: actually use synchronous transfers

[Jeff Epler]

this reduces a source of latency
---
 drivers/spi/spidev.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/drivers/spi/spidev.c b/drivers/spi/spidev.c
index dda7632..6e37c59 100644
--- a/drivers/spi/spidev.c
+++ b/drivers/spi/spidev.c
@@ -120,15 +120,9 @@ spidev_sync(struct spidev_data *spidev, struct spi_message *message)
 	if (spidev->spi == NULL)
 		status = -ESHUTDOWN;
 	else
-		status = spi_async(spidev->spi, message);
+		status = spi_sync(spidev->spi, message);
 	spin_unlock_irq(&spidev->spi_lock);
 
-	if (status == 0) {
-		wait_for_completion(&done);
-		status = message->status;
-		if (status == 0)
-			status = message->actual_length;
-	}
 	return status;
 }
 
-- 
2.0.1

[PATCH 4/4] spidev-s3c64xx: allocate dma channel at startup

[Jeff Epler]

it appears that acquire_dma introduces intolerable delays; the system
is fine with the dma channnels allocated all the time.
---
 drivers/spi/spi-s3c64xx.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/drivers/spi/spi-s3c64xx.c b/drivers/spi/spi-s3c64xx.c
index 92de1b5..e6a46c9 100644
--- a/drivers/spi/spi-s3c64xx.c
+++ b/drivers/spi/spi-s3c64xx.c
@@ -797,10 +797,6 @@ static int s3c64xx_spi_prepare_transfer(struct spi_master *spi)
 {
 	struct s3c64xx_spi_driver_data *sdd = spi_master_get_devdata(spi);
 
-	/* Acquire DMA channels */
-	while (!acquire_dma(sdd))
-		msleep(10);
-
 	pm_runtime_get_sync(&sdd->pdev->dev);
 
 	return 0;
@@ -810,10 +806,6 @@ static int s3c64xx_spi_unprepare_transfer(struct spi_master *spi)
 {
 	struct s3c64xx_spi_driver_data *sdd = spi_master_get_devdata(spi);
 
-	/* Free DMA channels */
-	sdd->ops->release(sdd->rx_dma.ch, &s3c64xx_spi_dma_client);
-	sdd->ops->release(sdd->tx_dma.ch, &s3c64xx_spi_dma_client);
-
 	pm_runtime_put(&sdd->pdev->dev);
 
 	return 0;
@@ -1344,6 +1336,9 @@ static int __init s3c64xx_spi_probe(struct platform_device *pdev)
 		goto err7;
 	}
 
+        // cannot fail(?)
+	acquire_dma(sdd);
+
 	writel(S3C64XX_SPI_INT_RX_OVERRUN_EN | S3C64XX_SPI_INT_RX_UNDERRUN_EN |
 	       S3C64XX_SPI_INT_TX_OVERRUN_EN | S3C64XX_SPI_INT_TX_UNDERRUN_EN,
 	       sdd->regs + S3C64XX_SPI_INT_EN);
@@ -1396,6 +1391,10 @@ static int s3c64xx_spi_remove(struct platform_device *pdev)
 
 	spi_unregister_master(master);
 
+	/* Free DMA channels */
+	sdd->ops->release(sdd->rx_dma.ch, &s3c64xx_spi_dma_client);
+	sdd->ops->release(sdd->tx_dma.ch, &s3c64xx_spi_dma_client);
+
 	writel(0, sdd->regs + S3C64XX_SPI_INT_EN);
 
 	free_irq(platform_get_irq(pdev, 0), sdd);
-- 
2.0.1

Re: [RFC 0/4] Improving SPI driver latency (vs v3.8.13.14-rt31)

[Philipp Lutz]

Hi Jeff,

thanks for your effort of investigating this issue.
I'll have a look if it can help me decreasing SPI latency with my ARM
based boards (Gumstix, BeagleBoneBlack).

Have you looked into setting a RTPRIO for the SPI worker threads
(spi_init_queue() in spi.c)? As far as I see hardly any SPI
hardware-dependent driver sets the "rt" property in the respective
spi_master struct, so that the main SPI driver (spi.c) will not use a RT
priority, which will be MAX_RT_PRIO - 1.

So far setting a medium RTPRIO for the SPI worker threads helped a lot
to reduce latency. Now I'm able to sustain a stable SPI transaction rate
of around 800 Hz.

Cheers
Phil


-------- Original Message --------
Subject: [RFC 0/4] Improving SPI driver latency (vs v3.8.13.14-rt31)
From: Jeff Epler <jepler@unpythonic.net>
To: linux-rt-users@vger.kernel.org
Date: 09/01/2014 04:30 PM
> I recently became interested in realtime access to an SPI device on the
> Odroid U3 platform, with a goal of running a repetitive task every 1ms
> that performs two SPI transactions. (for http://linuxcnc.org/
> interfacing to a http://mesanet.com/ "Anything I/O" card)
>
> Unfortunately, I found that there were frequent large latencies, some
> over 10ms, when using /dev/spidev.  This seems to be typical of others'
> experience using it (for instance, one can find threads discussing
> disappointing RT performance of SPI on the beaglebone and pandaboard; at
> least one raspberry pi project chose to implement a pure userspace SPI
> driver instead of using spidev)
>
> At all levels of the SPI stack, I found things that could be improved if
> lowest delays are the goal.  I doubt that in their current form these
> changes are suitable to be incorporated in preempt-rt, but I hope that 
> this might spur some discussion that would ultimately lead to better
> realtime performance of SPI in the preempt-rt kernel.
>
> As you may know, the kernel's spi driver stack consists of
>     spidev    - implementation of the /dev/spidev* device
>     spi       - hardware-independent kernel layer
>     spi-xyzzy - hardware-dependent driver for xyzzy hardware
>                 (s3c64xx in my device)
>
> I fixed causes of latency at each layer
>  * First, I eliminated per-ioctl memory allocations in spidev 
>  * Second, I made __spi_sync *actually* synchronous, rather than
>    being a wrapper over spi_async + wait_for_completion; and changed
>    spidev to use spi_sync
>  * Third, in the hardware-dependent code I moved DMA acquisition to
>    device initialization time rather than transaction time
>
> I did not quite achieve my goal of a 1ms repetitive rate yet, but with
> these changes I have run for 12+ hours at a rate of 3 transactions per
> 2ms with acceptable worst-case performance---under 250us for the biggest
> transaction, and 465us for all three (they have different sizes), with
> typical figures of more like 200us for all three transactions.  This is
> in contrast to the original performance, in which transactions taking
> over 10ms were seen multiple times per hour.  (12 hours is about 64
> million spi transations)
>
> (I changed from talking about 2 transactions to 3, because for an
> unrelated reason the communication in my program is currently divided
> into 3 SPI transactions when two would do)
>
> I know that 3.8 is by no means current, but 3.8.y is the default kernel
> shipped by hardkernel for their U3 devices so it was a rational version
> choice for me.  I did skim spi changes from version 3.8 to the present
> and didn't see anything that looked like it was directed at improving
> SPI latency, though the underlying code has probably changed enough over
> time that I assume my patches wouldn't actually apply at the tip of the
> latest and greatest branches.
>
> PS The fact that the first PREEMPT-RT kernel I built for the odroid
> worked and has basically good latency (until trying to talk to the
> hardware :-/) impressed the heck out of me.
>
> Jeff Epler (4):
>   spi: reenable sync SPI transfers
>   spidev: Avoid runtime memory allocations
>   spidev: actually use synchronous transfers
>   spidev-s3c64xx: allocate dma channel at startup
>
>  drivers/spi/spi-s3c64xx.c | 15 +++++++--------
>  drivers/spi/spi.c         | 22 ++++++++++++++++------
>  drivers/spi/spidev.c      | 43 ++++++++++++++++++++++++++++++-------------
>  3 files changed, 53 insertions(+), 27 deletions(-)
>

Re: [RFC 0/4] Improving SPI driver latency (vs v3.8.13.14-rt31)

[Jeff Epler]

On Mon, Sep 01, 2014 at 05:41:30PM +0200, Philipp Lutz wrote:
> Have you looked into setting a RTPRIO for the SPI worker threads
> (spi_init_queue() in spi.c)? As far as I see hardly any SPI
> hardware-dependent driver sets the "rt" property in the respective
> spi_master struct, so that the main SPI driver (spi.c) will not use a RT
> priority, which will be MAX_RT_PRIO - 1.

Sort of.

I was unaware of the "rt" property so I never tried setting it on the
kernel side.  However, I did run into the idea of using "chrt" on the
kernel tasks [spi1] and [irq/99-spi-s3c6].  This improved things
somewhat, but this plus eliminating DMA acquistion still left me with
some 800+us latencies that I didn't want.

(it may have also helped somewhat to to use cpuset to bind these tasks
to the same CPU where my realtime task runs--the U3 is a 4-core chip,
and linuxcnc binds to the last CPU by default)

Jeff

Re: [RFC 0/4] Improving SPI driver latency (vs v3.8.13.14-rt31)

[Harry van Haaren]

On Mon, Sep 1, 2014 at 7:13 PM, Jeff Epler <jepler@unpythonic.net> wrote:
> On Mon, Sep 01, 2014 at 05:41:30PM +0200, Philipp Lutz wrote:
>> Have you looked into setting a RTPRIO for the SPI worker threads
>> <snip>
> Sort of.
>
> I was unaware of the "rt" property so I never tried setting it on the
> kernel side.  However, I did run into the idea of using "chrt" on the
> kernel tasks [spi1] and [irq/99-spi-s3c6].  This improved things
> somewhat, but this plus eliminating DMA acquistion still left me with
> some 800+us latencies that I didn't want.

chrt changes the rtprio, its the same thing.
Prioritizing IRQ's and using chrt is discussed (with a focus on audio
soundcards) [1]. Although the page is outdated, it is the best resource
I'm aware of for practical advice / walktrough of using rtprio and such.

[1] http://subversion.ffado.org/wiki/IrqPriorities

Cheers, -Harry

www.openavproductions.com

Re: [RFC 0/4] Improving SPI driver latency (vs v3.8.13.14-rt31)

[Uwe Kleine-König]

Hello Jeff,

On Mon, Sep 01, 2014 at 09:30:31AM -0500, Jeff Epler wrote:
> Unfortunately, I found that there were frequent large latencies, some
> over 10ms, when using /dev/spidev.  This seems to be typical of others'
> experience using it (for instance, one can find threads discussing
> disappointing RT performance of SPI on the beaglebone and pandaboard; at
> least one raspberry pi project chose to implement a pure userspace SPI
> driver instead of using spidev)
> 
> At all levels of the SPI stack, I found things that could be improved if
> lowest delays are the goal.  I doubt that in their current form these
> changes are suitable to be incorporated in preempt-rt, but I hope that 
> this might spur some discussion that would ultimately lead to better
> realtime performance of SPI in the preempt-rt kernel.
I wonder why you didn't consider mainline as target for your patches
instead of preempt-rt. Also you might want to cc the spi maintainer on
this topic. I didn't check in detail what you did, but if it's useful
for others it would be a pity if it were not picked up because the
responsible people were not aware of your effort.

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html