* [PATCH RFCv6 2/2] misc: add CARMA DATA-FPGA Programmer support
From: Ira W. Snyder @ 2011-02-10 0:22 UTC (permalink / raw)
To: linuxppc-dev; +Cc: dmitry.torokhov, linux-kernel, Ira W. Snyder
In-Reply-To: <1297297376-19554-1-git-send-email-iws@ovro.caltech.edu>
This adds support for programming the data processing FPGAs on the OVRO
CARMA board. These FPGAs have a special programming sequence that
requires that we program the Freescale DMA engine, which is only
available inside the kernel.
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
---
drivers/misc/carma/Kconfig | 9 +
drivers/misc/carma/Makefile | 1 +
drivers/misc/carma/carma-fpga-program.c | 1084 +++++++++++++++++++++++++++++++
3 files changed, 1094 insertions(+), 0 deletions(-)
create mode 100644 drivers/misc/carma/carma-fpga-program.c
diff --git a/drivers/misc/carma/Kconfig b/drivers/misc/carma/Kconfig
index 4be183f..e57a9d3 100644
--- a/drivers/misc/carma/Kconfig
+++ b/drivers/misc/carma/Kconfig
@@ -7,3 +7,12 @@ config CARMA_FPGA
Say Y here to include support for communicating with the data
processing FPGAs on the OVRO CARMA board.
+config CARMA_FPGA_PROGRAM
+ tristate "CARMA DATA-FPGA Programmer"
+ depends on FSL_SOC && PPC_83xx && MEDIA_SUPPORT && HAS_DMA && FSL_DMA
+ select VIDEOBUF_DMA_SG
+ default n
+ help
+ Say Y here to include support for programming the data processing
+ FPGAs on the OVRO CARMA board.
+
diff --git a/drivers/misc/carma/Makefile b/drivers/misc/carma/Makefile
index 0b69fa7..ff36ac2 100644
--- a/drivers/misc/carma/Makefile
+++ b/drivers/misc/carma/Makefile
@@ -1 +1,2 @@
obj-$(CONFIG_CARMA_FPGA) += carma-fpga.o
+obj-$(CONFIG_CARMA_FPGA_PROGRAM) += carma-fpga-program.o
diff --git a/drivers/misc/carma/carma-fpga-program.c b/drivers/misc/carma/carma-fpga-program.c
new file mode 100644
index 0000000..ef16cb3
--- /dev/null
+++ b/drivers/misc/carma/carma-fpga-program.c
@@ -0,0 +1,1084 @@
+/*
+ * CARMA Board DATA-FPGA Programmer
+ *
+ * Copyright (c) 2009-2010 Ira W. Snyder <iws@ovro.caltech.edu>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation; either version 2 of the License, or (at your
+ * option) any later version.
+ */
+
+#include <linux/dma-mapping.h>
+#include <linux/of_platform.h>
+#include <linux/completion.h>
+#include <linux/miscdevice.h>
+#include <linux/dmaengine.h>
+#include <linux/interrupt.h>
+#include <linux/highmem.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/leds.h>
+#include <linux/slab.h>
+#include <linux/fs.h>
+#include <linux/io.h>
+
+#include <media/videobuf-dma-sg.h>
+
+/* MPC8349EMDS specific get_immrbase() */
+#include <sysdev/fsl_soc.h>
+
+static const char drv_name[] = "carma-fpga-program";
+
+/*
+ * Maximum firmware size
+ *
+ * 12849552 bytes for a CARMA Digitizer Board
+ * 18662880 bytes for a CARMA Correlator Board
+ */
+#define FW_SIZE_EP2S90 12849552
+#define FW_SIZE_EP2S130 18662880
+
+struct fpga_dev {
+ struct miscdevice miscdev;
+
+ /* Device Registers */
+ struct device *dev;
+ void __iomem *regs;
+ void __iomem *immr;
+
+ /* Freescale DMA Device */
+ struct dma_chan *chan;
+
+ /* Interrupts */
+ int irq, status;
+ struct completion completion;
+
+ /* FPGA Bitfile */
+ struct mutex lock;
+
+ struct videobuf_dmabuf vb;
+ bool vb_allocated;
+
+ /* max size and written bytes */
+ size_t fw_size;
+ size_t bytes;
+};
+
+/*
+ * FPGA Bitfile Helpers
+ */
+
+/**
+ * fpga_drop_firmware_data() - drop the bitfile image from memory
+ * @priv: the driver's private data structure
+ *
+ * LOCKING: must hold priv->lock
+ */
+static void fpga_drop_firmware_data(struct fpga_dev *priv)
+{
+ videobuf_dma_free(&priv->vb);
+ priv->vb_allocated = false;
+ priv->bytes = 0;
+}
+
+/*
+ * LED Trigger (could be a seperate module)
+ */
+
+/*
+ * NOTE: this whole thing does have the problem that whenever the led's are
+ * NOTE: first set to use the fpga trigger, they could be in the wrong state
+ */
+
+DEFINE_LED_TRIGGER(ledtrig_fpga);
+
+static void ledtrig_fpga_programmed(bool enabled)
+{
+ if (enabled)
+ led_trigger_event(ledtrig_fpga, LED_FULL);
+ else
+ led_trigger_event(ledtrig_fpga, LED_OFF);
+}
+
+/*
+ * FPGA Register Helpers
+ */
+
+/* Register Definitions */
+#define FPGA_CONFIG_CONTROL 0x40
+#define FPGA_CONFIG_STATUS 0x44
+#define FPGA_CONFIG_FIFO_SIZE 0x48
+#define FPGA_CONFIG_FIFO_USED 0x4C
+#define FPGA_CONFIG_TOTAL_BYTE_COUNT 0x50
+#define FPGA_CONFIG_CUR_BYTE_COUNT 0x54
+
+#define FPGA_FIFO_ADDRESS 0x3000
+
+static int fpga_fifo_size(void __iomem *regs)
+{
+ return ioread32be(regs + FPGA_CONFIG_FIFO_SIZE);
+}
+
+static int fpga_config_error(void __iomem *regs)
+{
+ return ioread32be(regs + FPGA_CONFIG_STATUS) & 0xFFFE;
+}
+
+static int fpga_fifo_empty(void __iomem *regs)
+{
+ return ioread32be(regs + FPGA_CONFIG_FIFO_USED) == 0;
+}
+
+static void fpga_fifo_write(void __iomem *regs, u32 val)
+{
+ iowrite32be(val, regs + FPGA_FIFO_ADDRESS);
+}
+
+static void fpga_set_byte_count(void __iomem *regs, u32 count)
+{
+ iowrite32be(count, regs + FPGA_CONFIG_TOTAL_BYTE_COUNT);
+}
+
+static void fpga_programmer_enable(struct fpga_dev *priv, bool dma)
+{
+ if (dma)
+ iowrite32be(0x5, priv->regs + FPGA_CONFIG_CONTROL);
+ else
+ iowrite32be(0x1, priv->regs + FPGA_CONFIG_CONTROL);
+}
+
+static void fpga_programmer_disable(struct fpga_dev *priv)
+{
+ iowrite32be(0x0, priv->regs + FPGA_CONFIG_CONTROL);
+}
+
+static void fpga_dump_registers(struct fpga_dev *priv)
+{
+ /* good status: do nothing */
+ if (priv->status == 0)
+ return;
+
+ /* Dump all status registers */
+ dev_err(priv->dev, "Configuration failed, dumping status registers\n");
+ dev_err(priv->dev, "Control: 0x%.8x\n", ioread32be(priv->regs + FPGA_CONFIG_CONTROL));
+ dev_err(priv->dev, "Status: 0x%.8x\n", ioread32be(priv->regs + FPGA_CONFIG_STATUS));
+ dev_err(priv->dev, "FIFO Size: 0x%.8x\n", ioread32be(priv->regs + FPGA_CONFIG_FIFO_SIZE));
+ dev_err(priv->dev, "FIFO Used: 0x%.8x\n", ioread32be(priv->regs + FPGA_CONFIG_FIFO_USED));
+ dev_err(priv->dev, "FIFO Total: 0x%.8x\n", ioread32be(priv->regs + FPGA_CONFIG_TOTAL_BYTE_COUNT));
+ dev_err(priv->dev, "FIFO Curr: 0x%.8x\n", ioread32be(priv->regs + FPGA_CONFIG_CUR_BYTE_COUNT));
+}
+
+/*
+ * FPGA Power Supply Code
+ */
+
+#define CTL_PWR_CONTROL 0x2006
+#define CTL_PWR_STATUS 0x200A
+#define CTL_PWR_FAIL 0x200B
+
+/*
+ * Determine if the FPGA power is good for all supplies
+ */
+static bool fpga_power_good(struct fpga_dev *priv)
+{
+ u8 val;
+
+ val = ioread8(priv->regs + CTL_PWR_STATUS);
+ if (val & 0x10)
+ return false;
+
+ return val == 0x0F;
+}
+
+/*
+ * Disable the FPGA power supplies
+ */
+static void fpga_disable_power_supplies(struct fpga_dev *priv)
+{
+ unsigned long start;
+ u8 val;
+
+ iowrite8(0x00, priv->regs + CTL_PWR_CONTROL);
+
+ /*
+ * Wait 500ms for the power rails to discharge
+ *
+ * Without this delay, the CTL-CPLD state machine can get into a
+ * state where it is waiting for the power-goods to assert, but they
+ * never do. This only happens when enabling and disabling the
+ * power sequencer very rapidly.
+ *
+ * The loop below will also wait for the power goods to de-assert,
+ * but testing has shown that they are always disabled by the time
+ * the sleep completes. However, omitting the sleep and only waiting
+ * for the power-goods to de-assert was not sufficient to ensure
+ * that the power sequencer would not wedge itself.
+ */
+ msleep(500);
+
+ start = jiffies;
+ while (time_before(jiffies, start + HZ)) {
+ val = ioread8(priv->regs + CTL_PWR_STATUS);
+ if (!(val & 0x0f))
+ break;
+
+ msleep(10);
+ }
+
+ val = ioread8(priv->regs + CTL_PWR_STATUS);
+ if (val & 0x0f) {
+ dev_err(priv->dev, "power disable failed: "
+ "power goods: status 0x%.2x\n", val);
+ }
+
+ if (val & 0x10) {
+ dev_err(priv->dev, "power disable failed: "
+ "alarm bit set: status 0x%.2x\n", val);
+ }
+}
+
+/**
+ * fpga_enable_power_supplies() - enable the DATA-FPGA power supplies
+ * @priv: the driver's private data structure
+ *
+ * Enable the DATA-FPGA power supplies, waiting up to 1 second for
+ * them to enable successfully.
+ *
+ * Returns 0 on success, -ERRNO otherwise
+ */
+static int fpga_enable_power_supplies(struct fpga_dev *priv)
+{
+ unsigned long start = jiffies;
+
+ if (fpga_power_good(priv)) {
+ dev_dbg(priv->dev, "power was already good\n");
+ return 0;
+ }
+
+ iowrite8(0x01, priv->regs + CTL_PWR_CONTROL);
+ while (time_before(jiffies, start + HZ)) {
+ if (fpga_power_good(priv))
+ return 0;
+
+ msleep(10);
+ }
+
+ return fpga_power_good(priv) ? 0 : -EBUSY;
+}
+
+/*
+ * Determine if the FPGA power supplies are all enabled
+ */
+static bool fpga_power_enabled(struct fpga_dev *priv)
+{
+ u8 val;
+
+ val = ioread8(priv->regs + CTL_PWR_CONTROL);
+ if (val & 0x01)
+ return true;
+
+ return false;
+}
+
+/*
+ * Determine if the FPGA's are programmed and running correctly
+ */
+static bool fpga_running(struct fpga_dev *priv)
+{
+ if (!fpga_power_good(priv))
+ return false;
+
+ /* Check the config done bit */
+ return ioread32be(priv->regs + FPGA_CONFIG_STATUS) & (1 << 18);
+}
+
+/*
+ * FPGA Programming Code
+ */
+
+/**
+ * fpga_program_block() - put a block of data into the programmer's FIFO
+ * @priv: the driver's private data structure
+ * @buf: the data to program
+ * @count: the length of data to program (must be a multiple of 4 bytes)
+ *
+ * Returns 0 on success, -ERRNO otherwise
+ */
+static int fpga_program_block(struct fpga_dev *priv, void *buf, size_t count)
+{
+ u32 *data = buf;
+ int size = fpga_fifo_size(priv->regs);
+ int i, len;
+ unsigned long timeout;
+
+ /* FIXME: BUG_ON instead */
+ WARN_ON_ONCE(count % 4 != 0);
+
+ while (count > 0) {
+
+ /* Get the size of the block to write (maximum is FIFO_SIZE) */
+ len = min_t(size_t, count, size);
+ timeout = jiffies + HZ / 4;
+
+ /* Write the block */
+ for (i = 0; i < len / 4; i++)
+ fpga_fifo_write(priv->regs, data[i]);
+
+ /* Update the amounts left */
+ count -= len;
+ data += len / 4;
+
+ /* Wait for the fifo to empty */
+ while (true) {
+
+ if (fpga_fifo_empty(priv->regs)) {
+ break;
+ } else {
+ dev_dbg(priv->dev, "Fifo not empty\n");
+ cpu_relax();
+ }
+
+ if (fpga_config_error(priv->regs)) {
+ dev_err(priv->dev, "Error detected\n");
+ return -EIO;
+ }
+
+ if (time_after(jiffies, timeout)) {
+ dev_err(priv->dev, "Fifo drain timeout\n");
+ return -ETIMEDOUT;
+ }
+
+ msleep(10);
+ }
+ }
+
+ return 0;
+}
+
+/**
+ * fpga_program_cpu() - program the DATA-FPGA's using the CPU
+ * @priv: the driver's private data structure
+ *
+ * This is useful when the DMA programming method fails. It is possible to
+ * wedge the Freescale DMA controller such that the DMA programming method
+ * always fails. This method has always succeeded.
+ *
+ * Returns 0 on success, -ERRNO otherwise
+ */
+static noinline int fpga_program_cpu(struct fpga_dev *priv)
+{
+ int ret;
+
+ /* Disable the programmer */
+ fpga_programmer_disable(priv);
+
+ /* Set the total byte count */
+ fpga_set_byte_count(priv->regs, priv->bytes);
+ dev_dbg(priv->dev, "total byte count %u bytes\n", priv->bytes);
+
+ /* Enable the controller for programming */
+ fpga_programmer_enable(priv, false);
+ dev_dbg(priv->dev, "enabled the controller\n");
+
+ /* Write each chunk of the FPGA bitfile to FPGA programmer */
+ ret = fpga_program_block(priv, priv->vb.vaddr, priv->bytes);
+ if (ret)
+ goto out_disable_controller;
+
+ /* Wait for the interrupt handler to notify us that programming finished */
+ ret = wait_for_completion_timeout(&priv->completion, 2 * HZ);
+ if (!ret) {
+ dev_err(priv->dev, "Timed out waiting for completion\n");
+ ret = -ETIMEDOUT;
+ goto out_disable_controller;
+ }
+
+ /* Retrieve the status from the interrupt handler */
+ ret = priv->status;
+
+out_disable_controller:
+ fpga_programmer_disable(priv);
+ return ret;
+}
+
+/**
+ * fpga_program_dma() - program the DATA-FPGA's using the DMA engine
+ * @priv: the driver's private data structure
+ *
+ * Program the DATA-FPGA's using the Freescale DMA engine. This requires that
+ * the engine is programmed such that the hardware DMA request lines can
+ * control the entire DMA transaction. The system controller FPGA then
+ * completely offloads the programming from the CPU.
+ *
+ * Returns 0 on success, -ERRNO otherwise
+ */
+static noinline int fpga_program_dma(struct fpga_dev *priv)
+{
+ struct videobuf_dmabuf *vb = &priv->vb;
+ struct dma_chan *chan = priv->chan;
+ struct dma_async_tx_descriptor *tx;
+ size_t num_pages, len, avail = 0;
+ struct dma_slave_config config;
+ struct scatterlist *sg;
+ struct sg_table table;
+ dma_cookie_t cookie;
+ int ret, i;
+
+ /* Disable the programmer */
+ fpga_programmer_disable(priv);
+
+ /* Allocate a scatterlist for the DMA destination */
+ num_pages = DIV_ROUND_UP(priv->bytes, 0x1000);
+ ret = sg_alloc_table(&table, num_pages, GFP_KERNEL);
+ if (ret) {
+ dev_err(priv->dev, "Unable to allocate dst scatterlist\n");
+ ret = -ENOMEM;
+ goto out_return;
+ }
+
+ /*
+ * This is an ugly hack
+ *
+ * We fill in a scatterlist as if it were mapped for DMA. This is
+ * necessary because there exists no better structure for this
+ * inside the kernel code.
+ *
+ * As an added bonus, we can use the DMAEngine API for all of this,
+ * rather than inventing another extremely similar API.
+ */
+ avail = priv->bytes;
+ for_each_sg(table.sgl, sg, num_pages, i) {
+ len = min_t(size_t, avail, 0x1000);
+ sg_dma_address(sg) = 0xf0003000;
+ sg_dma_len(sg) = len;
+
+ avail -= len;
+ }
+
+ /* Map the buffer for DMA */
+ ret = videobuf_dma_map(priv->dev, &priv->vb);
+ if (ret) {
+ dev_err(priv->dev, "Unable to map buffer for DMA\n");
+ goto out_free_table;
+ }
+
+ /*
+ * Configure the DMA channel to transfer FIFO_SIZE / 2 bytes per
+ * transaction, and then put it under external control
+ */
+ memset(&config, 0, sizeof(config));
+ config.direction = DMA_TO_DEVICE;
+ config.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
+ config.dst_maxburst = fpga_fifo_size(priv->regs) / 2 / 4;
+ ret = chan->device->device_control(chan, DMA_SLAVE_CONFIG,
+ (unsigned long)&config);
+ if (ret) {
+ dev_err(priv->dev, "DMA slave configuration failed\n");
+ goto out_dma_unmap;
+ }
+
+ ret = chan->device->device_control(chan, FSLDMA_EXTERNAL_START, 1);
+ if (ret) {
+ dev_err(priv->dev, "DMA external control setup failed\n");
+ goto out_dma_unmap;
+ }
+
+ /* setup and submit the DMA transaction */
+ tx = chan->device->device_prep_dma_sg(chan,
+ table.sgl, num_pages,
+ vb->sglist, vb->sglen, 0);
+ if (!tx) {
+ dev_err(priv->dev, "Unable to prep DMA transaction\n");
+ ret = -ENOMEM;
+ goto out_dma_unmap;
+ }
+
+ cookie = tx->tx_submit(tx);
+ if (dma_submit_error(cookie)) {
+ dev_err(priv->dev, "Unable to submit DMA transaction\n");
+ ret = -ENOMEM;
+ goto out_dma_unmap;
+ }
+
+ dma_async_memcpy_issue_pending(chan);
+
+ /* Set the total byte count */
+ fpga_set_byte_count(priv->regs, priv->bytes);
+ dev_dbg(priv->dev, "total byte count %u bytes\n", priv->bytes);
+
+ /* Enable the controller for DMA programming */
+ fpga_programmer_enable(priv, true);
+ dev_dbg(priv->dev, "enabled the controller\n");
+
+ /* Wait for the interrupt handler to notify us that programming finished */
+ ret = wait_for_completion_timeout(&priv->completion, 2 * HZ);
+ if (!ret) {
+ dev_err(priv->dev, "Timed out waiting for completion\n");
+ ret = -ETIMEDOUT;
+ goto out_disable_controller;
+ }
+
+ /* Retrieve the status from the interrupt handler */
+ ret = priv->status;
+
+out_disable_controller:
+ fpga_programmer_disable(priv);
+out_dma_unmap:
+ videobuf_dma_unmap(priv->dev, vb);
+out_free_table:
+ sg_free_table(&table);
+out_return:
+ return ret;
+}
+
+/*
+ * Interrupt Handling
+ */
+
+static irqreturn_t fpga_interrupt(int irq, void *dev_id)
+{
+ struct fpga_dev *priv = dev_id;
+
+ /* Save the status */
+ priv->status = fpga_config_error(priv->regs) ? -EIO : 0;
+ dev_dbg(priv->dev, "INTERRUPT status %d\n", priv->status);
+ fpga_dump_registers(priv);
+
+ /* Disabling the programmer clears the interrupt */
+ fpga_programmer_disable(priv);
+
+ /* Notify any waiters */
+ complete(&priv->completion);
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * SYSFS Helpers
+ */
+
+/**
+ * fpga_do_stop() - deconfigure (reset) the DATA-FPGA's
+ * @priv: the driver's private data structure
+ *
+ * LOCKING: must hold priv->lock
+ */
+static int fpga_do_stop(struct fpga_dev *priv)
+{
+ /* Set the led to unprogrammed */
+ ledtrig_fpga_programmed(false);
+
+ /* Pulse the config line to reset the FPGA's */
+ iowrite32be(0x3, priv->regs + FPGA_CONFIG_CONTROL);
+ iowrite32be(0x0, priv->regs + FPGA_CONFIG_CONTROL);
+
+ return 0;
+}
+
+static noinline int fpga_do_program(struct fpga_dev *priv)
+{
+ int ret;
+
+ if (priv->bytes != priv->fw_size) {
+ dev_err(priv->dev, "Incorrect bitfile size: got %zu bytes, "
+ "should be %zu bytes\n",
+ priv->bytes, priv->fw_size);
+ return -EINVAL;
+ }
+
+ if (!fpga_power_enabled(priv)) {
+ dev_err(priv->dev, "Power not enabled\n");
+ return -EINVAL;
+ }
+
+ if (!fpga_power_good(priv)) {
+ dev_err(priv->dev, "Power not good\n");
+ return -EINVAL;
+ }
+
+ /* Set the LED to unprogrammed */
+ ledtrig_fpga_programmed(false);
+
+ /* Try to program the FPGA's using DMA */
+ ret = fpga_program_dma(priv);
+
+ /* If DMA failed or doesn't exist, try with CPU */
+ if (ret) {
+ dev_warn(priv->dev, "Falling back to CPU programming\n");
+ ret = fpga_program_cpu(priv);
+ }
+
+ if (ret) {
+ dev_err(priv->dev, "Unable to program FPGA's\n");
+ return ret;
+ }
+
+ /* Drop the firmware bitfile from memory */
+ fpga_drop_firmware_data(priv);
+
+ dev_dbg(priv->dev, "FPGA programming successful\n");
+ ledtrig_fpga_programmed(true);
+
+ return 0;
+}
+
+/*
+ * File Operations
+ */
+
+static int fpga_open(struct inode *inode, struct file *filp)
+{
+ /*
+ * The miscdevice layer puts our struct miscdevice into the
+ * filp->private_data field. We use this to find our private
+ * data and then overwrite it with our own private structure.
+ */
+ struct fpga_dev *priv = container_of(filp->private_data,
+ struct fpga_dev, miscdev);
+ unsigned int nr_pages;
+ int ret;
+
+ /* We only allow one process at a time */
+ if (mutex_lock_interruptible(&priv->lock))
+ return -ERESTARTSYS;
+
+ filp->private_data = priv;
+
+ /* Truncation: drop any existing data */
+ if (filp->f_flags & O_TRUNC)
+ priv->bytes = 0;
+
+ /* Check if we have already allocated a buffer */
+ if (priv->vb_allocated)
+ return 0;
+
+ /* Allocate a buffer to hold enough data for the bitfile */
+ nr_pages = DIV_ROUND_UP(priv->fw_size, PAGE_SIZE);
+ ret = videobuf_dma_init_kernel(&priv->vb, DMA_TO_DEVICE, nr_pages);
+ if (ret) {
+ dev_err(priv->dev, "unable to allocate data buffer\n");
+ mutex_unlock(&priv->lock);
+ return ret;
+ }
+
+ priv->vb_allocated = true;
+ return 0;
+}
+
+static int fpga_release(struct inode *inode, struct file *filp)
+{
+ struct fpga_dev *priv = filp->private_data;
+
+ mutex_unlock(&priv->lock);
+ return 0;
+}
+
+static ssize_t fpga_write(struct file *filp, const char __user *buf,
+ size_t count, loff_t *f_pos)
+{
+ struct fpga_dev *priv = filp->private_data;
+
+ /* FPGA bitfiles have an exact size: disallow anything else */
+ if (priv->bytes >= priv->fw_size)
+ return -ENOSPC;
+
+ count = min_t(size_t, priv->fw_size - priv->bytes, count);
+ if (copy_from_user(priv->vb.vaddr + priv->bytes, buf, count))
+ return -EFAULT;
+
+ priv->bytes += count;
+ return count;
+}
+
+static ssize_t fpga_read(struct file *filp, char __user *buf, size_t count,
+ loff_t *f_pos)
+{
+ struct fpga_dev *priv = filp->private_data;
+
+ count = min_t(size_t, priv->bytes - *f_pos, count);
+ if (copy_to_user(buf, priv->vb.vaddr + *f_pos, count))
+ return -EFAULT;
+
+ *f_pos += count;
+ return count;
+}
+
+static loff_t fpga_llseek(struct file *filp, loff_t offset, int origin)
+{
+ struct fpga_dev *priv = filp->private_data;
+ loff_t newpos;
+
+ /* only read-only opens are allowed to seek */
+ if ((filp->f_flags & O_ACCMODE) != O_RDONLY)
+ return -EINVAL;
+
+ switch (origin) {
+ case SEEK_SET: /* seek relative to the beginning of the file */
+ newpos = offset;
+ break;
+ case SEEK_CUR: /* seek relative to current position in the file */
+ newpos = filp->f_pos + offset;
+ break;
+ case SEEK_END: /* seek relative to the end of the file */
+ newpos = priv->fw_size - offset;
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ /* check for sanity */
+ if (newpos > priv->fw_size)
+ return -EINVAL;
+
+ filp->f_pos = newpos;
+ return newpos;
+}
+
+static const struct file_operations fpga_fops = {
+ .open = fpga_open,
+ .release = fpga_release,
+ .write = fpga_write,
+ .read = fpga_read,
+ .llseek = fpga_llseek,
+};
+
+/*
+ * Device Attributes
+ */
+
+static ssize_t pfail_show(struct device *dev, struct device_attribute *attr,
+ char *buf)
+{
+ struct fpga_dev *priv = dev_get_drvdata(dev);
+ u8 val;
+
+ val = ioread8(priv->regs + CTL_PWR_FAIL);
+ return snprintf(buf, PAGE_SIZE, "0x%.2x\n", val);
+}
+
+static ssize_t pgood_show(struct device *dev, struct device_attribute *attr,
+ char *buf)
+{
+ struct fpga_dev *priv = dev_get_drvdata(dev);
+ return snprintf(buf, PAGE_SIZE, "%d\n", fpga_power_good(priv));
+}
+
+static ssize_t penable_show(struct device *dev, struct device_attribute *attr,
+ char *buf)
+{
+ struct fpga_dev *priv = dev_get_drvdata(dev);
+ return snprintf(buf, PAGE_SIZE, "%d\n", fpga_power_enabled(priv));
+}
+
+static ssize_t penable_store(struct device *dev, struct device_attribute *attr,
+ const char *buf, size_t count)
+{
+ struct fpga_dev *priv = dev_get_drvdata(dev);
+ unsigned long val;
+ int ret;
+
+ if (strict_strtoul(buf, 0, &val))
+ return -EINVAL;
+
+ if (val) {
+ ret = fpga_enable_power_supplies(priv);
+ if (ret)
+ return ret;
+ } else {
+ fpga_do_stop(priv);
+ fpga_disable_power_supplies(priv);
+ }
+
+ return count;
+}
+
+static ssize_t program_show(struct device *dev, struct device_attribute *attr,
+ char *buf)
+{
+ struct fpga_dev *priv = dev_get_drvdata(dev);
+ return snprintf(buf, PAGE_SIZE, "%d\n", fpga_running(priv));
+}
+
+static ssize_t program_store(struct device *dev, struct device_attribute *attr,
+ const char *buf, size_t count)
+{
+ struct fpga_dev *priv = dev_get_drvdata(dev);
+ unsigned long val;
+ int ret;
+
+ if (strict_strtoul(buf, 0, &val))
+ return -EINVAL;
+
+ /* We can't have an image writer and be programming simultaneously */
+ if (mutex_lock_interruptible(&priv->lock))
+ return -ERESTARTSYS;
+
+ /* Program or Reset the FPGA's */
+ ret = val ? fpga_do_program(priv) : fpga_do_stop(priv);
+ if (ret)
+ goto out_unlock;
+
+ /* Success */
+ ret = count;
+
+out_unlock:
+ mutex_unlock(&priv->lock);
+ return ret;
+}
+
+static DEVICE_ATTR(power_fail, S_IRUGO, pfail_show, NULL);
+static DEVICE_ATTR(power_good, S_IRUGO, pgood_show, NULL);
+static DEVICE_ATTR(power_enable, S_IRUGO | S_IWUGO, penable_show, penable_store);
+static DEVICE_ATTR(program, S_IRUGO | S_IWUGO, program_show, program_store);
+
+static struct attribute *fpga_attributes[] = {
+ &dev_attr_power_fail.attr,
+ &dev_attr_power_good.attr,
+ &dev_attr_power_enable.attr,
+ &dev_attr_program.attr,
+ NULL,
+};
+
+static const struct attribute_group fpga_attr_group = {
+ .attrs = fpga_attributes,
+};
+
+/*
+ * OpenFirmware Device Subsystem
+ */
+
+#define SYS_REG_VERSION 0x00
+#define SYS_REG_GEOGRAPHIC 0x10
+
+static bool dma_filter(struct dma_chan *chan, void *data)
+{
+ /*
+ * DMA Channel #0 is the only acceptable device
+ *
+ * This probably won't survive an unload/load cycle of the Freescale
+ * DMAEngine driver, but that won't be a problem
+ */
+ return chan->chan_id == 0 && chan->device->dev_id == 0;
+}
+
+static int fpga_of_remove(struct platform_device *op)
+{
+ struct fpga_dev *priv = dev_get_drvdata(&op->dev);
+ struct device *this_device = priv->miscdev.this_device;
+
+ sysfs_remove_group(&this_device->kobj, &fpga_attr_group);
+ misc_deregister(&priv->miscdev);
+
+ free_irq(priv->irq, priv);
+ iounmap(priv->immr);
+
+ fpga_disable_power_supplies(priv);
+ iounmap(priv->regs);
+
+ dma_release_channel(priv->chan);
+
+ /* Free any firmware image that has not been programmed */
+ fpga_drop_firmware_data(priv);
+
+ mutex_destroy(&priv->lock);
+ kfree(priv);
+
+ return 0;
+}
+
+static int fpga_of_probe(struct platform_device *op, const struct of_device_id *match)
+{
+ struct device_node *of_node = op->dev.of_node;
+ struct device *this_device;
+ struct fpga_dev *priv;
+ dma_cap_mask_t mask;
+ u32 ver;
+ int ret;
+
+ /* Allocate private data */
+ priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+ if (!priv) {
+ dev_err(&op->dev, "Unable to allocate private data\n");
+ ret = -ENOMEM;
+ goto out_return;
+ }
+
+ /* Setup the miscdevice */
+ priv->miscdev.minor = MISC_DYNAMIC_MINOR;
+ priv->miscdev.name = drv_name;
+ priv->miscdev.fops = &fpga_fops;
+
+ dev_set_drvdata(&op->dev, priv);
+ priv->dev = &op->dev;
+ mutex_init(&priv->lock);
+ init_completion(&priv->completion);
+ videobuf_dma_init(&priv->vb);
+
+ dev_set_drvdata(priv->dev, priv);
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_MEMCPY, mask);
+ dma_cap_set(DMA_INTERRUPT, mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ dma_cap_set(DMA_SG, mask);
+
+ /* Get control of DMA channel #0 */
+ priv->chan = dma_request_channel(mask, dma_filter, NULL);
+ if (!priv->chan) {
+ dev_err(&op->dev, "Unable to acquire DMA channel #0\n");
+ ret = -ENODEV;
+ goto out_free_priv;
+ }
+
+ /* Remap the registers for use */
+ priv->regs = of_iomap(of_node, 0);
+ if (!priv->regs) {
+ dev_err(&op->dev, "Unable to ioremap registers\n");
+ ret = -ENOMEM;
+ goto out_dma_release_channel;
+ }
+
+ /* Remap the IMMR for use */
+ priv->immr = ioremap(get_immrbase(), 0x100000);
+ if (!priv->immr) {
+ dev_err(&op->dev, "Unable to ioremap IMMR\n");
+ ret = -ENOMEM;
+ goto out_unmap_regs;
+ }
+
+ /*
+ * Check that external DMA is configured
+ *
+ * U-Boot does this for us, but we should check it and bail out if
+ * there is a problem. Failing to have this register setup correctly
+ * will cause the DMA controller to transfer a single cacheline
+ * worth of data, then wedge itself.
+ */
+ if ((ioread32be(priv->immr + 0x114) & 0xE00) != 0xE00) {
+ dev_err(&op->dev, "External DMA control not configured\n");
+ ret = -ENODEV;
+ goto out_unmap_immr;
+ }
+
+ /*
+ * Check the CTL-CPLD version
+ *
+ * This driver uses the CTL-CPLD DATA-FPGA power sequencer, and we
+ * don't want to run on any version of the CTL-CPLD that does not use
+ * a compatible register layout.
+ *
+ * v2: changed register layout, added power sequencer
+ * v3: added glitch filter on the i2c overcurrent/overtemp outputs
+ */
+ ver = ioread8(priv->regs + 0x2000);
+ if (ver != 0x02 && ver != 0x03) {
+ dev_err(&op->dev, "CTL-CPLD is not version 0x02 or 0x03!\n");
+ ret = -ENODEV;
+ goto out_unmap_immr;
+ }
+
+ /* Set the exact size that the firmware image should be */
+ ver = ioread32be(priv->regs + SYS_REG_VERSION);
+ priv->fw_size = (ver & (1 << 18)) ? FW_SIZE_EP2S130 : FW_SIZE_EP2S90;
+
+ /* Find the correct IRQ number */
+ priv->irq = irq_of_parse_and_map(of_node, 0);
+ if (priv->irq == NO_IRQ) {
+ dev_err(&op->dev, "Unable to find IRQ line\n");
+ ret = -ENODEV;
+ goto out_unmap_immr;
+ }
+
+ /* Request the IRQ */
+ ret = request_irq(priv->irq, fpga_interrupt, IRQF_SHARED, drv_name, priv);
+ if (ret) {
+ dev_err(&op->dev, "Unable to request IRQ %d\n", priv->irq);
+ ret = -ENODEV;
+ goto out_irq_dispose_mapping;
+ }
+
+ /* Reset and stop the FPGA's, just in case */
+ fpga_do_stop(priv);
+
+ /* Register the miscdevice */
+ ret = misc_register(&priv->miscdev);
+ if (ret) {
+ dev_err(&op->dev, "Unable to register miscdevice\n");
+ goto out_free_irq;
+ }
+
+ /* Create the sysfs files */
+ this_device = priv->miscdev.this_device;
+ dev_set_drvdata(this_device, priv);
+ ret = sysfs_create_group(&this_device->kobj, &fpga_attr_group);
+ if (ret) {
+ dev_err(&op->dev, "Unable to create sysfs files\n");
+ goto out_misc_deregister;
+ }
+
+ dev_info(priv->dev, "CARMA FPGA Programmer: %s rev%s with %s FPGAs\n",
+ (ver & (1 << 17)) ? "Correlator" : "Digitizer",
+ (ver & (1 << 16)) ? "B" : "A",
+ (ver & (1 << 18)) ? "EP2S130" : "EP2S90");
+
+ return 0;
+
+out_misc_deregister:
+ misc_deregister(&priv->miscdev);
+out_free_irq:
+ free_irq(priv->irq, priv);
+out_irq_dispose_mapping:
+ irq_dispose_mapping(priv->irq);
+out_unmap_immr:
+ iounmap(priv->immr);
+out_unmap_regs:
+ iounmap(priv->regs);
+out_dma_release_channel:
+ dma_release_channel(priv->chan);
+out_free_priv:
+ mutex_destroy(&priv->lock);
+ kfree(priv);
+out_return:
+ return ret;
+}
+
+static struct of_device_id fpga_of_match[] = {
+ { .compatible = "carma,fpga-programmer", },
+ {},
+};
+
+static struct of_platform_driver fpga_of_driver = {
+ .probe = fpga_of_probe,
+ .remove = fpga_of_remove,
+ .driver = {
+ .name = drv_name,
+ .of_match_table = fpga_of_match,
+ .owner = THIS_MODULE,
+ },
+};
+
+/*
+ * Module Init / Exit
+ */
+
+static int __init fpga_init(void)
+{
+ led_trigger_register_simple("fpga", &ledtrig_fpga);
+ return of_register_platform_driver(&fpga_of_driver);
+}
+
+static void __exit fpga_exit(void)
+{
+ of_unregister_platform_driver(&fpga_of_driver);
+ led_trigger_unregister_simple(ledtrig_fpga);
+}
+
+MODULE_AUTHOR("Ira W. Snyder <iws@ovro.caltech.edu>");
+MODULE_DESCRIPTION("CARMA Board DATA-FPGA Programmer");
+MODULE_LICENSE("GPL");
+
+module_init(fpga_init);
+module_exit(fpga_exit);
--
1.7.3.4
^ permalink raw reply related
* [PATCH RFCv6 1/2] misc: add CARMA DATA-FPGA Access Driver
From: Ira W. Snyder @ 2011-02-10 0:22 UTC (permalink / raw)
To: linuxppc-dev; +Cc: dmitry.torokhov, linux-kernel, Ira W. Snyder
In-Reply-To: <1297297376-19554-1-git-send-email-iws@ovro.caltech.edu>
This driver allows userspace to access the data processing FPGAs on the
OVRO CARMA board. It has two modes of operation:
1) random access
This allows users to poke any DATA-FPGA registers by using mmap to map
the address region directly into their memory map.
2) correlation dumping
When correlating, the DATA-FPGA's have special requirements for getting
the data out of their memory before the next correlation. This nominally
happens at 64Hz (every 15.625ms). If the data is not dumped before the
next correlation, data is lost.
The data dumping driver handles buffering up to 1 second worth of
correlation data from the FPGAs. This lowers the realtime scheduling
requirements for the userspace process reading the device.
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
---
drivers/misc/Kconfig | 1 +
drivers/misc/Makefile | 1 +
drivers/misc/carma/Kconfig | 9 +
drivers/misc/carma/Makefile | 1 +
drivers/misc/carma/carma-fpga.c | 1407 +++++++++++++++++++++++++++++++++++++++
5 files changed, 1419 insertions(+), 0 deletions(-)
create mode 100644 drivers/misc/carma/Kconfig
create mode 100644 drivers/misc/carma/Makefile
create mode 100644 drivers/misc/carma/carma-fpga.c
diff --git a/drivers/misc/Kconfig b/drivers/misc/Kconfig
index cc8e49d..93cf1e6 100644
--- a/drivers/misc/Kconfig
+++ b/drivers/misc/Kconfig
@@ -457,5 +457,6 @@ source "drivers/misc/eeprom/Kconfig"
source "drivers/misc/cb710/Kconfig"
source "drivers/misc/iwmc3200top/Kconfig"
source "drivers/misc/ti-st/Kconfig"
+source "drivers/misc/carma/Kconfig"
endif # MISC_DEVICES
diff --git a/drivers/misc/Makefile b/drivers/misc/Makefile
index 98009cc..2c1610e 100644
--- a/drivers/misc/Makefile
+++ b/drivers/misc/Makefile
@@ -42,3 +42,4 @@ obj-$(CONFIG_ARM_CHARLCD) += arm-charlcd.o
obj-$(CONFIG_PCH_PHUB) += pch_phub.o
obj-y += ti-st/
obj-$(CONFIG_AB8500_PWM) += ab8500-pwm.o
+obj-y += carma/
diff --git a/drivers/misc/carma/Kconfig b/drivers/misc/carma/Kconfig
new file mode 100644
index 0000000..4be183f
--- /dev/null
+++ b/drivers/misc/carma/Kconfig
@@ -0,0 +1,9 @@
+config CARMA_FPGA
+ tristate "CARMA DATA-FPGA Access Driver"
+ depends on FSL_SOC && PPC_83xx && MEDIA_SUPPORT && HAS_DMA && FSL_DMA
+ select VIDEOBUF_DMA_SG
+ default n
+ help
+ Say Y here to include support for communicating with the data
+ processing FPGAs on the OVRO CARMA board.
+
diff --git a/drivers/misc/carma/Makefile b/drivers/misc/carma/Makefile
new file mode 100644
index 0000000..0b69fa7
--- /dev/null
+++ b/drivers/misc/carma/Makefile
@@ -0,0 +1 @@
+obj-$(CONFIG_CARMA_FPGA) += carma-fpga.o
diff --git a/drivers/misc/carma/carma-fpga.c b/drivers/misc/carma/carma-fpga.c
new file mode 100644
index 0000000..be40a07
--- /dev/null
+++ b/drivers/misc/carma/carma-fpga.c
@@ -0,0 +1,1407 @@
+/*
+ * CARMA DATA-FPGA Access Driver
+ *
+ * Copyright (c) 2009-2011 Ira W. Snyder <iws@ovro.caltech.edu>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation; either version 2 of the License, or (at your
+ * option) any later version.
+ */
+
+/*
+ * FPGA Memory Dump Format
+ *
+ * FPGA #0 control registers (32 x 32-bit words)
+ * FPGA #1 control registers (32 x 32-bit words)
+ * FPGA #2 control registers (32 x 32-bit words)
+ * FPGA #3 control registers (32 x 32-bit words)
+ * SYSFPGA control registers (32 x 32-bit words)
+ * FPGA #0 correlation array (NUM_CORL0 correlation blocks)
+ * FPGA #1 correlation array (NUM_CORL1 correlation blocks)
+ * FPGA #2 correlation array (NUM_CORL2 correlation blocks)
+ * FPGA #3 correlation array (NUM_CORL3 correlation blocks)
+ *
+ * Each correlation array consists of:
+ *
+ * Correlation Data (2 x NUM_LAGSn x 32-bit words)
+ * Pipeline Metadata (2 x NUM_METAn x 32-bit words)
+ * Quantization Counters (2 x NUM_QCNTn x 32-bit words)
+ *
+ * The NUM_CORLn, NUM_LAGSn, NUM_METAn, and NUM_QCNTn values come from
+ * the FPGA configuration registers. They do not change once the FPGA's
+ * have been programmed, they only change on re-programming.
+ */
+
+/*
+ * Basic Description:
+ *
+ * This driver is used to capture correlation spectra off of the four data
+ * processing FPGAs. The FPGAs are often reprogrammed at runtime, therefore
+ * this driver supports dynamic enable/disable of capture while the device
+ * remains open.
+ *
+ * The nominal capture rate is 64Hz (every 15.625ms). To facilitate this fast
+ * capture rate, all buffers are pre-allocated to avoid any potentially long
+ * running memory allocations while capturing.
+ *
+ * There are two lists and one pointer which are used to keep track of the
+ * different states of data buffers.
+ *
+ * 1) free list
+ * This list holds all empty data buffers which are ready to receive data.
+ *
+ * 2) inflight pointer
+ * This pointer holds the currently inflight data buffer. This buffer is having
+ * data copied into it by the DMA engine.
+ *
+ * 3) used list
+ * This list holds data buffers which have been filled, and are waiting to be
+ * read by userspace.
+ *
+ * All buffers start life on the free list, then move successively to the
+ * inflight pointer, and then to the used list. After they have been read by
+ * userspace, they are moved back to the free list. The cycle repeats as long
+ * as necessary.
+ *
+ * It should be noted that all buffers are mapped and ready for DMA when they
+ * are on any of the three lists. They are only unmapped when they are in the
+ * process of being read by userspace.
+ */
+
+/*
+ * Notes on the IRQ masking scheme:
+ *
+ * The IRQ masking scheme here is different than most other hardware. The only
+ * way for the DATA-FPGAs to detect if the kernel has taken too long to copy
+ * the data is if the status registers are not cleared before the next
+ * correlation data dump is ready.
+ *
+ * The interrupt line is connected to the status registers, such that when they
+ * are cleared, the interrupt is de-asserted. Therein lies our problem. We need
+ * to schedule a long-running DMA operation and return from the interrupt
+ * handler quickly, but we cannot clear the status registers.
+ *
+ * To handle this, the system controller FPGA has the capability to connect the
+ * interrupt line to a user-controlled GPIO pin. This pin is driven high
+ * (unasserted) and left that way. To mask the interrupt, we change the
+ * interrupt source to the GPIO pin. Tada, we hid the interrupt. :)
+ */
+
+#include <linux/of_platform.h>
+#include <linux/dma-mapping.h>
+#include <linux/miscdevice.h>
+#include <linux/interrupt.h>
+#include <linux/dmaengine.h>
+#include <linux/seq_file.h>
+#include <linux/highmem.h>
+#include <linux/debugfs.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/poll.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/io.h>
+
+#include <media/videobuf-dma-sg.h>
+
+/* system controller registers */
+#define SYS_IRQ_SOURCE_CTL 0x24
+#define SYS_IRQ_OUTPUT_EN 0x28
+#define SYS_IRQ_OUTPUT_DATA 0x2C
+#define SYS_IRQ_INPUT_DATA 0x30
+#define SYS_FPGA_CONFIG_STATUS 0x44
+
+/* GPIO IRQ line assignment */
+#define IRQ_CORL_DONE 0x10
+
+/* FPGA registers */
+#define MMAP_REG_VERSION 0x00
+#define MMAP_REG_CORL_CONF1 0x08
+#define MMAP_REG_CORL_CONF2 0x0C
+#define MMAP_REG_STATUS 0x48
+
+#define SYS_FPGA_BLOCK 0xF0000000
+
+#define DATA_FPGA_START 0x400000
+#define DATA_FPGA_SIZE 0x80000
+
+static const char drv_name[] = "carma-fpga";
+
+#define NUM_FPGA 4
+
+#define MIN_DATA_BUFS 8
+#define MAX_DATA_BUFS 64
+
+struct fpga_info {
+ unsigned int num_lag_ram;
+ unsigned int blk_size;
+};
+
+struct data_buf {
+ struct list_head entry;
+ struct videobuf_dmabuf vb;
+ size_t size;
+};
+
+struct fpga_device {
+ struct miscdevice miscdev;
+ struct device *dev;
+ struct mutex mutex;
+
+ /* FPGA registers and information */
+ struct fpga_info info[NUM_FPGA];
+ void __iomem *regs;
+ int irq;
+
+ /* FPGA Physical Address/Size Information */
+ resource_size_t phys_addr;
+ size_t phys_size;
+
+ /* DMA structures */
+ struct sg_table corl_table;
+ unsigned int corl_nents;
+ struct dma_chan *chan;
+
+ /* Protection for all members below */
+ spinlock_t lock;
+
+ /* Device enable/disable flag */
+ bool enabled;
+
+ /* Correlation data buffers */
+ wait_queue_head_t wait;
+ struct list_head free;
+ struct list_head used;
+ struct data_buf *inflight;
+
+ /* Information about data buffers */
+ unsigned int num_dropped;
+ unsigned int num_buffers;
+ size_t bufsize;
+ struct dentry *dbg_entry;
+};
+
+struct fpga_reader {
+ struct fpga_device *priv;
+ struct data_buf *buf;
+ off_t buf_start;
+};
+
+/*
+ * Data Buffer Allocation Helpers
+ */
+
+/**
+ * data_free_buffer() - free a single data buffer and all allocated memory
+ * @buf: the buffer to free
+ *
+ * This will free all of the pages allocated to the given data buffer, and
+ * then free the structure itself
+ */
+static void data_free_buffer(struct data_buf *buf)
+{
+ /* It is ok to free a NULL buffer */
+ if (!buf)
+ return;
+
+ /* free all memory */
+ videobuf_dma_free(&buf->vb);
+ kfree(buf);
+}
+
+/**
+ * data_alloc_buffer() - allocate and fill a data buffer with pages
+ * @bytes: the number of bytes required
+ *
+ * This allocates all space needed for a data buffer. It must be mapped before
+ * use in a DMA transaction using videobuf_dma_map().
+ *
+ * Returns NULL on failure
+ */
+static struct data_buf *data_alloc_buffer(const size_t bytes)
+{
+ unsigned int nr_pages;
+ struct data_buf *buf;
+ int ret;
+
+ /* calculate the number of pages necessary */
+ nr_pages = DIV_ROUND_UP(bytes, PAGE_SIZE);
+
+ /* allocate the buffer structure */
+ buf = kzalloc(sizeof(*buf), GFP_KERNEL);
+ if (!buf)
+ goto out_return;
+
+ /* initialize internal fields */
+ INIT_LIST_HEAD(&buf->entry);
+ buf->size = bytes;
+
+ /* allocate the videobuf */
+ videobuf_dma_init(&buf->vb);
+ ret = videobuf_dma_init_kernel(&buf->vb, DMA_FROM_DEVICE, nr_pages);
+ if (ret)
+ goto out_free_buf;
+
+ return buf;
+
+out_free_buf:
+ kfree(buf);
+out_return:
+ return NULL;
+}
+
+/**
+ * data_free_buffers() - free all allocated buffers
+ * @priv: the driver's private data structure
+ *
+ * Free all buffers allocated by the driver (except those currently in the
+ * process of being read by userspace).
+ *
+ * LOCKING: must hold dev->mutex
+ * CONTEXT: user
+ */
+static void data_free_buffers(struct fpga_device *priv)
+{
+ struct data_buf *buf, *tmp;
+
+ /* the device should be stopped, no DMA in progress */
+ BUG_ON(priv->inflight != NULL);
+
+ list_for_each_entry_safe(buf, tmp, &priv->free, entry) {
+ list_del_init(&buf->entry);
+ videobuf_dma_unmap(priv->dev, &buf->vb);
+ data_free_buffer(buf);
+ }
+
+ list_for_each_entry_safe(buf, tmp, &priv->used, entry) {
+ list_del_init(&buf->entry);
+ videobuf_dma_unmap(priv->dev, &buf->vb);
+ data_free_buffer(buf);
+ }
+
+ priv->num_buffers = 0;
+ priv->bufsize = 0;
+}
+
+/**
+ * data_alloc_buffers() - allocate 1 seconds worth of data buffers
+ * @priv: the driver's private data structure
+ *
+ * Allocate enough buffers for a whole second worth of data
+ *
+ * This routine will attempt to degrade nicely by succeeding even if a full
+ * second worth of data buffers could not be allocated, as long as a minimum
+ * number were allocated. In this case, it will print a message to the kernel
+ * log.
+ *
+ * The device must not be modifying any lists when this is called.
+ *
+ * CONTEXT: user
+ * LOCKING: must hold dev->mutex
+ *
+ * Returns 0 on success, -ERRNO otherwise
+ */
+static int data_alloc_buffers(struct fpga_device *priv)
+{
+ struct data_buf *buf;
+ int i, ret;
+
+ for (i = 0; i < MAX_DATA_BUFS; i++) {
+
+ /* allocate a buffer */
+ buf = data_alloc_buffer(priv->bufsize);
+ if (!buf)
+ break;
+
+ /* map it for DMA */
+ ret = videobuf_dma_map(priv->dev, &buf->vb);
+ if (ret) {
+ data_free_buffer(buf);
+ break;
+ }
+
+ /* add it to the list of free buffers */
+ list_add_tail(&buf->entry, &priv->free);
+ priv->num_buffers++;
+ }
+
+ /* Make sure we allocated the minimum required number of buffers */
+ if (priv->num_buffers < MIN_DATA_BUFS) {
+ dev_err(priv->dev, "Unable to allocate enough data buffers\n");
+ data_free_buffers(priv);
+ return -ENOMEM;
+ }
+
+ /* Warn if we are running in a degraded state, but do not fail */
+ if (priv->num_buffers < MAX_DATA_BUFS) {
+ dev_warn(priv->dev,
+ "Unable to allocate %d buffers, using %d buffers instead\n",
+ MAX_DATA_BUFS, i);
+ }
+
+ return 0;
+}
+
+/*
+ * DMA Operations Helpers
+ */
+
+/**
+ * fpga_start_addr() - get the physical address a DATA-FPGA
+ * @priv: the driver's private data structure
+ * @fpga: the DATA-FPGA number (zero based)
+ */
+static dma_addr_t fpga_start_addr(struct fpga_device *priv, unsigned int fpga)
+{
+ return priv->phys_addr + 0x400000 + (0x80000 * fpga);
+}
+
+/**
+ * fpga_block_addr() - get the physical address of a correlation data block
+ * @priv: the driver's private data structure
+ * @fpga: the DATA-FPGA number (zero based)
+ * @blknum: the correlation block number (zero based)
+ */
+static dma_addr_t fpga_block_addr(struct fpga_device *priv, unsigned int fpga,
+ unsigned int blknum)
+{
+ return fpga_start_addr(priv, fpga) + (0x10000 * (1 + blknum));
+}
+
+#define REG_BLOCK_SIZE (32 * 4)
+
+/**
+ * data_setup_corl_table() - create the scatterlist for correlation dumps
+ * @priv: the driver's private data structure
+ *
+ * Create the scatterlist for transferring a correlation dump from the
+ * DATA FPGAs. This structure will be reused for each buffer than needs
+ * to be filled with correlation data.
+ *
+ * Returns 0 on success, -ERRNO otherwise
+ */
+static int data_setup_corl_table(struct fpga_device *priv)
+{
+ struct sg_table *table = &priv->corl_table;
+ struct scatterlist *sg;
+ struct fpga_info *info;
+ int i, j, ret;
+
+ /* Calculate the number of entries needed */
+ priv->corl_nents = (1 + NUM_FPGA) * REG_BLOCK_SIZE;
+ for (i = 0; i < NUM_FPGA; i++)
+ priv->corl_nents += priv->info[i].num_lag_ram;
+
+ /* Allocate the scatterlist table */
+ ret = sg_alloc_table(table, priv->corl_nents, GFP_KERNEL);
+ if (ret) {
+ dev_err(priv->dev, "unable to allocate DMA table\n");
+ return ret;
+ }
+
+ /* Add the DATA FPGA registers to the scatterlist */
+ sg = table->sgl;
+ for (i = 0; i < NUM_FPGA; i++) {
+ sg_dma_address(sg) = fpga_start_addr(priv, i);
+ sg_dma_len(sg) = REG_BLOCK_SIZE;
+ sg = sg_next(sg);
+ }
+
+ /* Add the SYS-FPGA registers to the scatterlist */
+ sg_dma_address(sg) = SYS_FPGA_BLOCK;
+ sg_dma_len(sg) = REG_BLOCK_SIZE;
+ sg = sg_next(sg);
+
+ /* Add the FPGA correlation data blocks to the scatterlist */
+ for (i = 0; i < NUM_FPGA; i++) {
+ info = &priv->info[i];
+ for (j = 0; j < info->num_lag_ram; j++) {
+ sg_dma_address(sg) = fpga_block_addr(priv, i, j);
+ sg_dma_len(sg) = info->blk_size;
+ sg = sg_next(sg);
+ }
+ }
+
+ /*
+ * All physical addresses and lengths are present in the structure
+ * now. It can be reused for every FPGA DATA interrupt
+ */
+ return 0;
+}
+
+/*
+ * FPGA Register Access Helpers
+ */
+
+static void fpga_write_reg(struct fpga_device *priv, unsigned int fpga,
+ unsigned int reg, u32 val)
+{
+ const int fpga_start = DATA_FPGA_START + (fpga * DATA_FPGA_SIZE);
+ iowrite32be(val, priv->regs + fpga_start + reg);
+}
+
+static u32 fpga_read_reg(struct fpga_device *priv, unsigned int fpga,
+ unsigned int reg)
+{
+ const int fpga_start = DATA_FPGA_START + (fpga * DATA_FPGA_SIZE);
+ return ioread32be(priv->regs + fpga_start + reg);
+}
+
+/**
+ * data_calculate_bufsize() - calculate the data buffer size required
+ * @priv: the driver's private data structure
+ *
+ * Calculate the total buffer size needed to hold a single block
+ * of correlation data
+ *
+ * CONTEXT: user
+ *
+ * Returns 0 on success, -ERRNO otherwise
+ */
+static int data_calculate_bufsize(struct fpga_device *priv)
+{
+ u32 num_corl, num_lags, num_meta, num_qcnt, num_pack;
+ u32 conf1, conf2, version;
+ u32 num_lag_ram, blk_size;
+ int i;
+
+ /* Each buffer starts with the 5 FPGA register areas */
+ priv->bufsize = (1 + NUM_FPGA) * REG_BLOCK_SIZE;
+
+ /* Read and store the configuration data for each FPGA */
+ for (i = 0; i < NUM_FPGA; i++) {
+ version = fpga_read_reg(priv, i, MMAP_REG_VERSION);
+ conf1 = fpga_read_reg(priv, i, MMAP_REG_CORL_CONF1);
+ conf2 = fpga_read_reg(priv, i, MMAP_REG_CORL_CONF2);
+
+ /* minor version 2 and later */
+ if ((version & 0x000000FF) >= 2) {
+ num_corl = (conf1 & 0x000000F0) >> 4;
+ num_pack = (conf1 & 0x00000F00) >> 8;
+ num_lags = (conf1 & 0x00FFF000) >> 12;
+ num_meta = (conf1 & 0x7F000000) >> 24;
+ num_qcnt = (conf2 & 0x00000FFF) >> 0;
+ } else {
+ num_corl = (conf1 & 0x000000F0) >> 4;
+ num_pack = 1; /* implied */
+ num_lags = (conf1 & 0x000FFF00) >> 8;
+ num_meta = (conf1 & 0x7FF00000) >> 20;
+ num_qcnt = (conf2 & 0x00000FFF) >> 0;
+ }
+
+ num_lag_ram = (num_corl + num_pack - 1) / num_pack;
+ blk_size = ((num_pack * num_lags) + num_meta + num_qcnt) * 8;
+
+ priv->info[i].num_lag_ram = num_lag_ram;
+ priv->info[i].blk_size = blk_size;
+ priv->bufsize += num_lag_ram * blk_size;
+
+ dev_dbg(priv->dev, "FPGA %d NUM_CORL: %d\n", i, num_corl);
+ dev_dbg(priv->dev, "FPGA %d NUM_PACK: %d\n", i, num_pack);
+ dev_dbg(priv->dev, "FPGA %d NUM_LAGS: %d\n", i, num_lags);
+ dev_dbg(priv->dev, "FPGA %d NUM_META: %d\n", i, num_meta);
+ dev_dbg(priv->dev, "FPGA %d NUM_QCNT: %d\n", i, num_qcnt);
+ dev_dbg(priv->dev, "FPGA %d BLK_SIZE: %d\n", i, blk_size);
+ }
+
+ dev_dbg(priv->dev, "TOTAL BUFFER SIZE: %zu bytes\n", priv->bufsize);
+ return 0;
+}
+
+/*
+ * Interrupt Handling
+ */
+
+/**
+ * data_disable_interrupts() - stop the device from generating interrupts
+ * @priv: the driver's private data structure
+ *
+ * Hide interrupts by switching to GPIO interrupt source
+ *
+ * LOCKING: must hold dev->lock
+ */
+static void data_disable_interrupts(struct fpga_device *priv)
+{
+ /* hide the interrupt by switching the IRQ driver to GPIO */
+ iowrite32be(0x2F, priv->regs + SYS_IRQ_SOURCE_CTL);
+}
+
+/**
+ * data_enable_interrupts() - allow the device to generate interrupts
+ * @priv: the driver's private data structure
+ *
+ * Unhide interrupts by switching to the FPGA interrupt source. At the
+ * same time, clear the DATA-FPGA status registers.
+ *
+ * LOCKING: must hold dev->lock
+ */
+static void data_enable_interrupts(struct fpga_device *priv)
+{
+ /* clear the actual FPGA corl_done interrupt */
+ fpga_write_reg(priv, 0, MMAP_REG_STATUS, 0x0);
+ fpga_write_reg(priv, 1, MMAP_REG_STATUS, 0x0);
+ fpga_write_reg(priv, 2, MMAP_REG_STATUS, 0x0);
+ fpga_write_reg(priv, 3, MMAP_REG_STATUS, 0x0);
+
+ /* flush the writes */
+ fpga_read_reg(priv, 0, MMAP_REG_STATUS);
+
+ /* switch back to the external interrupt source */
+ iowrite32be(0x3F, priv->regs + SYS_IRQ_SOURCE_CTL);
+}
+
+/**
+ * data_dma_cb() - DMAEngine callback for DMA completion
+ * @data: the driver's private data structure
+ *
+ * Complete a DMA transfer from the DATA-FPGA's
+ *
+ * This is called via the DMA callback mechanism, and will handle moving the
+ * completed DMA transaction to the used list, and then wake any processes
+ * waiting for new data
+ *
+ * CONTEXT: any, softirq expected
+ */
+static void data_dma_cb(void *data)
+{
+ struct fpga_device *priv = data;
+ unsigned long flags;
+
+ spin_lock_irqsave(&priv->lock, flags);
+
+ /* If there is no inflight buffer, we've got a bug */
+ BUG_ON(priv->inflight == NULL);
+
+ /* Move the inflight buffer onto the used list */
+ list_move_tail(&priv->inflight->entry, &priv->used);
+ priv->inflight = NULL;
+
+ /* clear the FPGA status and re-enable interrupts */
+ data_enable_interrupts(priv);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+
+ /*
+ * We've changed both the inflight and used lists, so we need
+ * to wake up any processes that are blocking for those events
+ */
+ wake_up(&priv->wait);
+}
+
+/**
+ * data_submit_dma() - prepare and submit the required DMA to fill a buffer
+ * @priv: the driver's private data structure
+ * @buf: the data buffer
+ *
+ * Prepare and submit the necessary DMA transactions to fill a correlation
+ * data buffer.
+ *
+ * LOCKING: must hold dev->lock
+ * CONTEXT: hardirq only
+ *
+ * Returns 0 on success, -ERRNO otherwise
+ */
+static int data_submit_dma(struct fpga_device *priv, struct data_buf *buf)
+{
+ struct scatterlist *dst_sg, *src_sg;
+ unsigned int dst_nents, src_nents;
+ struct dma_chan *chan = priv->chan;
+ struct dma_async_tx_descriptor *tx;
+ dma_cookie_t cookie;
+ dma_addr_t dst, src;
+
+ dst_sg = buf->vb.sglist;
+ dst_nents = buf->vb.sglen;
+
+ src_sg = priv->corl_table.sgl;
+ src_nents = priv->corl_nents;
+
+ /*
+ * All buffers passed to this function should be ready and mapped
+ * for DMA already. Therefore, we don't need to do anything except
+ * submit it to the Freescale DMA Engine for processing
+ */
+
+ /* setup the scatterlist to scatterlist transfer */
+ tx = chan->device->device_prep_dma_sg(chan,
+ dst_sg, dst_nents,
+ src_sg, src_nents,
+ 0);
+ if (!tx) {
+ dev_err(priv->dev, "unable to prep scatterlist DMA\n");
+ return -ENOMEM;
+ }
+
+ /* submit the transaction to the DMA controller */
+ cookie = tx->tx_submit(tx);
+ if (dma_submit_error(cookie)) {
+ dev_err(priv->dev, "unable to submit scatterlist DMA\n");
+ return -ENOMEM;
+ }
+
+ /* Prepare the re-read of the SYS-FPGA block */
+ dst = sg_dma_address(dst_sg) + (NUM_FPGA * REG_BLOCK_SIZE);
+ src = SYS_FPGA_BLOCK;
+ tx = chan->device->device_prep_dma_memcpy(chan, dst, src,
+ REG_BLOCK_SIZE,
+ DMA_PREP_INTERRUPT);
+ if (!tx) {
+ dev_err(priv->dev, "unable to prep SYS-FPGA DMA\n");
+ return -ENOMEM;
+ }
+
+ /* Setup the callback */
+ tx->callback = data_dma_cb;
+ tx->callback_param = priv;
+
+ /* submit the transaction to the DMA controller */
+ cookie = tx->tx_submit(tx);
+ if (dma_submit_error(cookie)) {
+ dev_err(priv->dev, "unable to submit SYS-FPGA DMA\n");
+ return -ENOMEM;
+ }
+
+ return 0;
+}
+
+#define CORL_DONE 0x1
+#define CORL_ERR 0x2
+
+static irqreturn_t data_irq(int irq, void *dev_id)
+{
+ struct fpga_device *priv = dev_id;
+ bool submitted = false;
+ struct data_buf *buf;
+ u32 status;
+ int i;
+
+ /* detect spurious interrupts via FPGA status */
+ for (i = 0; i < 4; i++) {
+ status = fpga_read_reg(priv, i, MMAP_REG_STATUS);
+ if (!(status & (CORL_DONE | CORL_ERR))) {
+ dev_err(priv->dev, "spurious irq detected (FPGA)\n");
+ return IRQ_NONE;
+ }
+ }
+
+ /* detect spurious interrupts via raw IRQ pin readback */
+ status = ioread32be(priv->regs + SYS_IRQ_INPUT_DATA);
+ if (status & IRQ_CORL_DONE) {
+ dev_err(priv->dev, "spurious irq detected (IRQ)\n");
+ return IRQ_NONE;
+ }
+
+ spin_lock(&priv->lock);
+
+ /* hide the interrupt by switching the IRQ driver to GPIO */
+ data_disable_interrupts(priv);
+
+ /* If there are no free buffers, drop this data */
+ if (list_empty(&priv->free)) {
+ priv->num_dropped++;
+ goto out;
+ }
+
+ buf = list_first_entry(&priv->free, struct data_buf, entry);
+ list_del_init(&buf->entry);
+ BUG_ON(buf->size != priv->bufsize);
+
+ /* Submit a DMA transfer to get the correlation data */
+ if (data_submit_dma(priv, buf)) {
+ dev_err(priv->dev, "Unable to setup DMA transfer\n");
+ list_move_tail(&buf->entry, &priv->free);
+ goto out;
+ }
+
+ /* Save the buffer for the DMA callback */
+ priv->inflight = buf;
+ submitted = true;
+
+ /* Start the DMA Engine */
+ dma_async_memcpy_issue_pending(priv->chan);
+
+out:
+ /* If no DMA was submitted, re-enable interrupts */
+ if (!submitted)
+ data_enable_interrupts(priv);
+
+ spin_unlock(&priv->lock);
+ return IRQ_HANDLED;
+}
+
+/*
+ * Realtime Device Enable Helpers
+ */
+
+/**
+ * data_device_enable() - enable the device for buffered dumping
+ * @priv: the driver's private data structure
+ *
+ * Enable the device for buffered dumping. Allocates buffers and hooks up
+ * the interrupt handler. When this finishes, data will come pouring in.
+ *
+ * LOCKING: must hold dev->mutex
+ * CONTEXT: user context only
+ *
+ * Returns 0 on success, -ERRNO otherwise
+ */
+static int data_device_enable(struct fpga_device *priv)
+{
+ u32 val;
+ int ret;
+
+ /* multiple enables are safe: they do nothing */
+ if (priv->enabled)
+ return 0;
+
+ /* check that the FPGAs are programmed */
+ val = ioread32be(priv->regs + SYS_FPGA_CONFIG_STATUS);
+ if (!(val & (1 << 18))) {
+ dev_err(priv->dev, "DATA-FPGAs are not enabled\n");
+ return -ENODATA;
+ }
+
+ /* read the FPGAs to calculate the buffer size */
+ ret = data_calculate_bufsize(priv);
+ if (ret) {
+ dev_err(priv->dev, "unable to calculate buffer size\n");
+ goto out_error;
+ }
+
+ /* allocate the correlation data buffers */
+ ret = data_alloc_buffers(priv);
+ if (ret) {
+ dev_err(priv->dev, "unable to allocate buffers\n");
+ goto out_error;
+ }
+
+ /* setup the source scatterlist for dumping correlation data */
+ ret = data_setup_corl_table(priv);
+ if (ret) {
+ dev_err(priv->dev, "unable to setup correlation DMA table\n");
+ goto out_error;
+ }
+
+ /* hookup the irq handler */
+ ret = request_irq(priv->irq, data_irq, IRQF_SHARED, drv_name, priv);
+ if (ret) {
+ dev_err(priv->dev, "unable to request IRQ handler\n");
+ goto out_error;
+ }
+
+ /* switch to the external FPGA IRQ line */
+ data_enable_interrupts(priv);
+
+ /* success, we're enabled */
+ priv->enabled = true;
+ return 0;
+
+out_error:
+ sg_free_table(&priv->corl_table);
+ priv->corl_nents = 0;
+
+ data_free_buffers(priv);
+ return ret;
+}
+
+/**
+ * data_device_disable() - disable the device for buffered dumping
+ * @priv: the driver's private data structure
+ *
+ * Disable the device for buffered dumping. Stops new DMA transactions from
+ * being generated, waits for all outstanding DMA to complete, and then frees
+ * all buffers.
+ *
+ * LOCKING: must hold dev->mutex
+ * CONTEXT: user only
+ *
+ * Returns 0 on success, -ERRNO otherwise
+ */
+static int data_device_disable(struct fpga_device *priv)
+{
+ int ret;
+
+ /* allow multiple disable */
+ if (!priv->enabled)
+ return 0;
+
+ /* switch to the internal GPIO IRQ line */
+ data_disable_interrupts(priv);
+
+ /* unhook the irq handler */
+ free_irq(priv->irq, priv);
+
+ /*
+ * wait for all outstanding DMA to complete
+ *
+ * Device interrupts are disabled, therefore another buffer cannot
+ * be marked inflight.
+ */
+ ret = wait_event_interruptible(priv->wait, priv->inflight == NULL);
+ if (ret)
+ return ret;
+
+ /* free the correlation table */
+ sg_free_table(&priv->corl_table);
+ priv->corl_nents = 0;
+
+ /*
+ * We are taking the spinlock not to protect priv->enabled, but instead
+ * to make sure that there are no readers in the process of altering
+ * the free or used lists while we are setting this flag.
+ */
+ spin_lock_irq(&priv->lock);
+ priv->enabled = false;
+ spin_unlock_irq(&priv->lock);
+
+ /* free all buffers: the free and used lists are not being changed */
+ data_free_buffers(priv);
+ return 0;
+}
+
+/*
+ * DEBUGFS Interface
+ */
+#ifdef CONFIG_DEBUG_FS
+
+/*
+ * Count the number of entries in the given list
+ */
+static unsigned int list_num_entries(struct list_head *list)
+{
+ struct list_head *entry;
+ unsigned int ret = 0;
+
+ list_for_each(entry, list)
+ ret++;
+
+ return ret;
+}
+
+static int data_debug_show(struct seq_file *f, void *offset)
+{
+ struct fpga_device *priv = f->private;
+ int ret;
+
+ /*
+ * Lock the mutex first, so that we get an accurate value for enable
+ * Lock the spinlock next, to get accurate list counts
+ */
+ ret = mutex_lock_interruptible(&priv->mutex);
+ if (ret)
+ return ret;
+
+ spin_lock_irq(&priv->lock);
+
+ seq_printf(f, "enabled: %d\n", priv->enabled);
+ seq_printf(f, "bufsize: %d\n", priv->bufsize);
+ seq_printf(f, "num_buffers: %d\n", priv->num_buffers);
+ seq_printf(f, "num_free: %d\n", list_num_entries(&priv->free));
+ seq_printf(f, "inflight: %d\n", priv->inflight != NULL);
+ seq_printf(f, "num_used: %d\n", list_num_entries(&priv->used));
+ seq_printf(f, "num_dropped: %d\n", priv->num_dropped);
+
+ spin_unlock_irq(&priv->lock);
+ mutex_unlock(&priv->mutex);
+ return 0;
+}
+
+static int data_debug_open(struct inode *inode, struct file *file)
+{
+ return single_open(file, data_debug_show, inode->i_private);
+}
+
+static const struct file_operations data_debug_fops = {
+ .owner = THIS_MODULE,
+ .open = data_debug_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+};
+
+static int data_debugfs_init(struct fpga_device *priv)
+{
+ priv->dbg_entry = debugfs_create_file(drv_name, S_IRUGO, NULL, priv,
+ &data_debug_fops);
+ if (IS_ERR(priv->dbg_entry))
+ return PTR_ERR(priv->dbg_entry);
+
+ return 0;
+}
+
+static void data_debugfs_exit(struct fpga_device *priv)
+{
+ debugfs_remove(priv->dbg_entry);
+}
+
+#else
+
+static inline int data_debugfs_init(struct fpga_device *priv)
+{
+ return 0;
+}
+
+static inline void data_debugfs_exit(struct fpga_device *priv)
+{
+}
+
+#endif /* CONFIG_DEBUG_FS */
+
+/*
+ * SYSFS Attributes
+ */
+
+static ssize_t data_en_show(struct device *dev, struct device_attribute *attr,
+ char *buf)
+{
+ struct fpga_device *priv = dev_get_drvdata(dev);
+ return snprintf(buf, PAGE_SIZE, "%u\n", priv->enabled);
+}
+
+static ssize_t data_en_set(struct device *dev, struct device_attribute *attr,
+ const char *buf, size_t count)
+{
+ struct fpga_device *priv = dev_get_drvdata(dev);
+ unsigned long enable;
+ int ret;
+
+ ret = strict_strtoul(buf, 0, &enable);
+ if (ret) {
+ dev_err(priv->dev, "unable to parse enable input\n");
+ return -EINVAL;
+ }
+
+ ret = mutex_lock_interruptible(&priv->mutex);
+ if (ret)
+ return ret;
+
+ if (enable)
+ ret = data_device_enable(priv);
+ else
+ ret = data_device_disable(priv);
+
+ if (ret) {
+ dev_err(priv->dev, "device %s failed\n",
+ enable ? "enable" : "disable");
+ count = ret;
+ goto out_unlock;
+ }
+
+out_unlock:
+ mutex_unlock(&priv->mutex);
+ return count;
+}
+
+static DEVICE_ATTR(enable, S_IWUSR | S_IRUGO, data_en_show, data_en_set);
+
+static struct attribute *data_sysfs_attrs[] = {
+ &dev_attr_enable.attr,
+ NULL,
+};
+
+static const struct attribute_group rt_sysfs_attr_group = {
+ .attrs = data_sysfs_attrs,
+};
+
+/*
+ * FPGA Realtime Data Character Device
+ */
+
+static int data_open(struct inode *inode, struct file *filp)
+{
+ /*
+ * The miscdevice layer puts our struct miscdevice into the
+ * filp->private_data field. We use this to find our private
+ * data and then overwrite it with our own private structure.
+ */
+ struct fpga_device *priv = container_of(filp->private_data,
+ struct fpga_device, miscdev);
+ struct fpga_reader *reader;
+ int ret;
+
+ /* allocate private data */
+ reader = kzalloc(sizeof(*reader), GFP_KERNEL);
+ if (!reader)
+ return -ENOMEM;
+
+ reader->priv = priv;
+ reader->buf = NULL;
+
+ filp->private_data = reader;
+ ret = nonseekable_open(inode, filp);
+ if (ret) {
+ dev_err(priv->dev, "nonseekable-open failed\n");
+ kfree(reader);
+ return ret;
+ }
+
+ return 0;
+}
+
+static int data_release(struct inode *inode, struct file *filp)
+{
+ struct fpga_reader *reader = filp->private_data;
+
+ /* free the per-reader structure */
+ data_free_buffer(reader->buf);
+ kfree(reader);
+ filp->private_data = NULL;
+ return 0;
+}
+
+static ssize_t data_read(struct file *filp, char __user *ubuf, size_t count,
+ loff_t *f_pos)
+{
+ struct fpga_reader *reader = filp->private_data;
+ struct fpga_device *priv = reader->priv;
+ struct list_head *used = &priv->used;
+ struct data_buf *dbuf;
+ size_t avail;
+ void *data;
+ int ret;
+
+ /* check if we already have a partial buffer */
+ if (reader->buf) {
+ dbuf = reader->buf;
+ goto have_buffer;
+ }
+
+ spin_lock_irq(&priv->lock);
+
+ /* Block until there is at least one buffer on the used list */
+ while (list_empty(used)) {
+ spin_unlock_irq(&priv->lock);
+
+ if (filp->f_flags & O_NONBLOCK)
+ return -EAGAIN;
+
+ ret = wait_event_interruptible(priv->wait, !list_empty(used));
+ if (ret)
+ return ret;
+
+ spin_lock_irq(&priv->lock);
+ }
+
+ /* Grab the first buffer off of the used list */
+ dbuf = list_first_entry(used, struct data_buf, entry);
+ list_del_init(&dbuf->entry);
+
+ spin_unlock_irq(&priv->lock);
+
+ /* Buffers are always mapped: unmap it */
+ videobuf_dma_unmap(priv->dev, &dbuf->vb);
+
+ /* save the buffer for later */
+ reader->buf = dbuf;
+ reader->buf_start = 0;
+
+have_buffer:
+ /* Get the number of bytes available */
+ avail = dbuf->size - reader->buf_start;
+ data = dbuf->vb.vaddr + reader->buf_start;
+
+ /* Get the number of bytes we can transfer */
+ count = min(count, avail);
+
+ /* Copy the data to the userspace buffer */
+ if (copy_to_user(ubuf, data, count))
+ return -EFAULT;
+
+ /* Update the amount of available space */
+ avail -= count;
+
+ /*
+ * If there is still some data available, save the buffer for the
+ * next userspace call to read() and return
+ */
+ if (avail > 0) {
+ reader->buf_start += count;
+ reader->buf = dbuf;
+ return count;
+ }
+
+ /*
+ * Get the buffer ready to be reused for DMA
+ *
+ * If it fails, we pretend that the read never happed and return
+ * -EFAULT to userspace. The read will be retried.
+ */
+ ret = videobuf_dma_map(priv->dev, &dbuf->vb);
+ if (ret) {
+ dev_err(priv->dev, "unable to remap buffer for DMA\n");
+ return -EFAULT;
+ }
+
+ /* Lock against concurrent enable/disable */
+ spin_lock_irq(&priv->lock);
+
+ /* the reader is finished with this buffer */
+ reader->buf = NULL;
+
+ /*
+ * One of two things has happened, the device is disabled, or the
+ * device has been reconfigured underneath us. In either case, we
+ * should just throw away the buffer.
+ */
+ if (!priv->enabled || dbuf->size != priv->bufsize) {
+ videobuf_dma_unmap(priv->dev, &dbuf->vb);
+ data_free_buffer(dbuf);
+ goto out_unlock;
+ }
+
+ /* The buffer is safe to reuse, so add it back to the free list */
+ list_add_tail(&dbuf->entry, &priv->free);
+
+out_unlock:
+ spin_unlock_irq(&priv->lock);
+ return count;
+}
+
+static unsigned int data_poll(struct file *filp, struct poll_table_struct *tbl)
+{
+ struct fpga_reader *reader = filp->private_data;
+ struct fpga_device *priv = reader->priv;
+ unsigned int mask = 0;
+
+ poll_wait(filp, &priv->wait, tbl);
+
+ if (!list_empty(&priv->used))
+ mask |= POLLIN | POLLRDNORM;
+
+ return mask;
+}
+
+static int data_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+ struct fpga_reader *reader = filp->private_data;
+ struct fpga_device *priv = reader->priv;
+ unsigned long offset, vsize, psize, addr;
+
+ /* VMA properties */
+ offset = vma->vm_pgoff << PAGE_SHIFT;
+ vsize = vma->vm_end - vma->vm_start;
+ psize = priv->phys_size - offset;
+ addr = (priv->phys_addr + offset) >> PAGE_SHIFT;
+
+ /* Check against the FPGA region's physical memory size */
+ if (vsize > psize) {
+ dev_err(priv->dev, "requested mmap mapping too large\n");
+ return -EINVAL;
+ }
+
+ /* IO memory (stop cacheing) */
+ vma->vm_flags |= VM_IO | VM_RESERVED;
+ vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+
+ return io_remap_pfn_range(vma, vma->vm_start, addr, vsize,
+ vma->vm_page_prot);
+}
+
+static const struct file_operations data_fops = {
+ .owner = THIS_MODULE,
+ .open = data_open,
+ .release = data_release,
+ .read = data_read,
+ .poll = data_poll,
+ .mmap = data_mmap,
+ .llseek = no_llseek,
+};
+
+/*
+ * OpenFirmware Device Subsystem
+ */
+
+static bool dma_filter(struct dma_chan *chan, void *data)
+{
+ /*
+ * DMA Channel #0 is used for the FPGA Programmer, so ignore it
+ *
+ * This probably won't survive an unload/load cycle of the Freescale
+ * DMAEngine driver, but that won't be a problem
+ */
+ if (chan->chan_id == 0 && chan->device->dev_id == 0)
+ return false;
+
+ return true;
+}
+
+static int data_of_probe(struct platform_device *op,
+ const struct of_device_id *match)
+{
+ struct device_node *of_node = op->dev.of_node;
+ struct device *this_device;
+ struct fpga_device *priv;
+ struct resource res;
+ dma_cap_mask_t mask;
+ int ret;
+
+ /* Allocate private data */
+ priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+ if (!priv) {
+ dev_err(&op->dev, "Unable to allocate device private data\n");
+ ret = -ENOMEM;
+ goto out_return;
+ }
+
+ dev_set_drvdata(&op->dev, priv);
+ priv->dev = &op->dev;
+
+ /* Setup the misc device */
+ priv->miscdev.minor = MISC_DYNAMIC_MINOR;
+ priv->miscdev.name = drv_name;
+ priv->miscdev.fops = &data_fops;
+
+ /* Get the physical address of the FPGA registers */
+ ret = of_address_to_resource(of_node, 0, &res);
+ if (ret) {
+ dev_err(&op->dev, "Unable to find FPGA physical address\n");
+ ret = -ENODEV;
+ goto out_free_priv;
+ }
+
+ priv->phys_addr = res.start;
+ priv->phys_size = resource_size(&res);
+
+ /* ioremap the registers for use */
+ priv->regs = of_iomap(of_node, 0);
+ if (!priv->regs) {
+ dev_err(&op->dev, "Unable to ioremap registers\n");
+ ret = -ENOMEM;
+ goto out_free_priv;
+ }
+
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_MEMCPY, mask);
+ dma_cap_set(DMA_INTERRUPT, mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ dma_cap_set(DMA_SG, mask);
+
+ /* Request a DMA channel */
+ priv->chan = dma_request_channel(mask, dma_filter, NULL);
+ if (!priv->chan) {
+ dev_err(&op->dev, "Unable to request DMA channel\n");
+ ret = -ENODEV;
+ goto out_unmap_regs;
+ }
+
+ /* Find the correct IRQ number */
+ priv->irq = irq_of_parse_and_map(of_node, 0);
+ if (priv->irq == NO_IRQ) {
+ dev_err(&op->dev, "Unable to find IRQ line\n");
+ ret = -ENODEV;
+ goto out_release_dma;
+ }
+
+ dev_set_drvdata(priv->dev, priv);
+ mutex_init(&priv->mutex);
+ spin_lock_init(&priv->lock);
+ INIT_LIST_HEAD(&priv->free);
+ INIT_LIST_HEAD(&priv->used);
+ init_waitqueue_head(&priv->wait);
+
+ /* Drive the GPIO for FPGA IRQ high (no interrupt) */
+ iowrite32be(IRQ_CORL_DONE, priv->regs + SYS_IRQ_OUTPUT_DATA);
+
+ /* Register the miscdevice */
+ ret = misc_register(&priv->miscdev);
+ if (ret) {
+ dev_err(&op->dev, "Unable to register miscdevice\n");
+ goto out_irq_dispose_mapping;
+ }
+
+ /* Create the debugfs files */
+ ret = data_debugfs_init(priv);
+ if (ret) {
+ dev_err(&op->dev, "Unable to create debugfs files\n");
+ goto out_misc_deregister;
+ }
+
+ /* Create the sysfs files */
+ this_device = priv->miscdev.this_device;
+ dev_set_drvdata(this_device, priv);
+ ret = sysfs_create_group(&this_device->kobj, &rt_sysfs_attr_group);
+ if (ret) {
+ dev_err(&op->dev, "Unable to create sysfs files\n");
+ goto out_data_debugfs_exit;
+ }
+
+ dev_info(&op->dev, "CARMA FPGA Realtime Data Driver Loaded\n");
+ return 0;
+
+out_data_debugfs_exit:
+ data_debugfs_exit(priv);
+out_misc_deregister:
+ misc_deregister(&priv->miscdev);
+out_irq_dispose_mapping:
+ irq_dispose_mapping(priv->irq);
+out_release_dma:
+ dma_release_channel(priv->chan);
+out_unmap_regs:
+ iounmap(priv->regs);
+out_free_priv:
+ mutex_destroy(&priv->mutex);
+ kfree(priv);
+out_return:
+ return ret;
+}
+
+static int data_of_remove(struct platform_device *op)
+{
+ struct fpga_device *priv = dev_get_drvdata(&op->dev);
+ struct device *this_device = priv->miscdev.this_device;
+
+ /* remove all sysfs files, now the device cannot be re-enabled */
+ sysfs_remove_group(&this_device->kobj, &rt_sysfs_attr_group);
+
+ /* remove all debugfs files */
+ data_debugfs_exit(priv);
+
+ /* disable the device from generating data */
+ data_device_disable(priv);
+
+ misc_deregister(&priv->miscdev);
+ irq_dispose_mapping(priv->irq);
+ dma_release_channel(priv->chan);
+ iounmap(priv->regs);
+ mutex_destroy(&priv->mutex);
+ kfree(priv);
+
+ return 0;
+}
+
+static struct of_device_id data_of_match[] = {
+ { .compatible = "carma,carma-fpga", },
+ {},
+};
+
+static struct of_platform_driver data_of_driver = {
+ .probe = data_of_probe,
+ .remove = data_of_remove,
+ .driver = {
+ .name = drv_name,
+ .of_match_table = data_of_match,
+ .owner = THIS_MODULE,
+ },
+};
+
+/*
+ * Module Init / Exit
+ */
+
+static int __init data_init(void)
+{
+ return of_register_platform_driver(&data_of_driver);
+}
+
+static void __exit data_exit(void)
+{
+ of_unregister_platform_driver(&data_of_driver);
+}
+
+MODULE_AUTHOR("Ira W. Snyder <iws@ovro.caltech.edu>");
+MODULE_DESCRIPTION("CARMA DATA-FPGA Access Driver");
+MODULE_LICENSE("GPL");
+
+module_init(data_init);
+module_exit(data_exit);
--
1.7.3.4
^ permalink raw reply related
* [PATCH RFCv6 0/2] CARMA Board Support
From: Ira W. Snyder @ 2011-02-10 0:22 UTC (permalink / raw)
To: linuxppc-dev; +Cc: dmitry.torokhov, linux-kernel, Ira W. Snyder
Hello everyone,
This is the sixth posting of these drivers, taking into account comments from
earlier postings. I would appreciate as much review as you can offer.
RFCv5 -> RFCv6:
- change locking in several functions
- use list_move_tail() to simplify code
- remove unused helper functions
RFCv4 -> RFCv5:
- remove unecessary locking per review comments
- do not clobber return values from *_interruptible()
- explicitly track buffer DMA mapping
- use #defines instead of raw hex addresses
- change enable sysfs attribute to root-writeable only
RFCv3 -> RFCv4:
- updates for DATA-FPGA version 2
RFCv2 -> RFCv3:
- use miscdevice framework (removing the carma class)
- add bitfile readback capability to the programmer
RFCv1 -> RFCv2:
- change comments to kerneldoc format
- Kconfig improvements
- use the videobuf_dma_sg API in the programmer
- updates for Freescale DMAEngine DMA_SLAVE API changes
KNOWN ISSUES:
- untested with a setup that can generate interrupts (will get access soon)
- does not handle runtime "unbind"
Information about the CARMA board:
The CARMA board is essentially an MPC8349EA MDS reference design with a
1GHz ADC and 4 high powered data processing FPGAs connected to the local
bus. It is all packed into a compact PCI form factor. It is used at the
Owens Valley Radio Observatory as the main component in the correlator
system.
For board information, see:
http://www.mmarray.org/~dwh/carma_board/index.html
For DATA-FPGA register layout, see:
http://www.mmarray.org/memos/carma_memo46.pdf
These drivers are the necessary pieces to get the data processing FPGAs
working and producing data. Despite the fact that the hardware is custom
and we are the only users, I'd still like to get the drivers upstream.
Several people have suggested that this is possible.
Some further patches will be forthcoming. I have a driver for the LED
subsystem and the PPS subsystem. The LED register layout is expected to
change soon, so I won't post the driver until that is finished. The PPS
driver will be posted seperately from this patch series; it is very
generic.
Thanks to everyone who has provided comments on earlier versions!
Ira W. Snyder (2):
misc: add CARMA DATA-FPGA Access Driver
misc: add CARMA DATA-FPGA Programmer support
drivers/misc/Kconfig | 1 +
drivers/misc/Makefile | 1 +
drivers/misc/carma/Kconfig | 18 +
drivers/misc/carma/Makefile | 2 +
drivers/misc/carma/carma-fpga-program.c | 1084 ++++++++++++++++++++++++
drivers/misc/carma/carma-fpga.c | 1407 +++++++++++++++++++++++++++++++
6 files changed, 2513 insertions(+), 0 deletions(-)
create mode 100644 drivers/misc/carma/Kconfig
create mode 100644 drivers/misc/carma/Makefile
create mode 100644 drivers/misc/carma/carma-fpga-program.c
create mode 100644 drivers/misc/carma/carma-fpga.c
--
1.7.3.4
^ permalink raw reply
* Re: [PATCH 1/2] misc: add CARMA DATA-FPGA Access Driver
From: Ira W. Snyder @ 2011-02-10 0:10 UTC (permalink / raw)
To: Dmitry Torokhov; +Cc: linuxppc-dev, linux-kernel
In-Reply-To: <20110209234231.GA5707@core.coreip.homeip.net>
On Wed, Feb 09, 2011 at 03:42:31PM -0800, Dmitry Torokhov wrote:
> On Wed, Feb 09, 2011 at 03:35:45PM -0800, Ira W. Snyder wrote:
> > On Wed, Feb 09, 2011 at 10:27:40AM -0800, Dmitry Torokhov wrote:
> >
> > [ snip stuff I've already fixed in the next version ]
> >
> > > >
> > > > The requirement is that the device stay open during reconfiguration.
> > > > This provides for that. Readers just block for as long as the device is
> > > > not producing data.
> > >
> > > OK, you still need to make sure you do not touch free/used buffer while
> > > device is disabled. Also, you need to kick readers if you unbind the
> > > driver, so maybe a new flag priv->exists should be introduced and
> > > checked.
> > >
> >
> > I don't understand what you mean by "kick readers if you unbind the
> > driver". The kernel automatically increases the refcount on a module
> > when a process is using the module. This shows up in the "Used by"
> > column of lsmod's output.
> >
> > The kernel will not let you rmmod a module with a non-zero refcount. You
> > cannot get into the situation where you have rmmod'ed the module and a
> > reader is still blocking in read()/poll().
>
> However you can still unbind the driver from the device by writing into
> driver's sysfs 'unbind' attribute.
>
> See drivers/base/bus.c::driver_unbind().
>
I was completely unaware of that "feature". I hunch that many drivers
are incapable of dealing with an unbind while they are still open.
Matter of fact, I don't see how this can EVER be safe. The driver core
automatically calls the data_of_remove() routine while there are still
blocked readers. This kfree()s the private data structure, which
contains the suggested priv->exists flag. What happens if the memory
allocator re-allocates that memory to a different driver before the
reader process is woken up to check the priv->exists flag?
The only way to solve this is to count the number of open()s and
close()s, and block the unbind until all users have close()d the device.
Thanks,
Ira
^ permalink raw reply
* Re: [PATCH 1/2] misc: add CARMA DATA-FPGA Access Driver
From: Dmitry Torokhov @ 2011-02-09 23:42 UTC (permalink / raw)
To: Ira W. Snyder; +Cc: linuxppc-dev, linux-kernel
In-Reply-To: <20110209233544.GA5303@ovro.caltech.edu>
On Wed, Feb 09, 2011 at 03:35:45PM -0800, Ira W. Snyder wrote:
> On Wed, Feb 09, 2011 at 10:27:40AM -0800, Dmitry Torokhov wrote:
>
> [ snip stuff I've already fixed in the next version ]
>
> > >
> > > The requirement is that the device stay open during reconfiguration.
> > > This provides for that. Readers just block for as long as the device is
> > > not producing data.
> >
> > OK, you still need to make sure you do not touch free/used buffer while
> > device is disabled. Also, you need to kick readers if you unbind the
> > driver, so maybe a new flag priv->exists should be introduced and
> > checked.
> >
>
> I don't understand what you mean by "kick readers if you unbind the
> driver". The kernel automatically increases the refcount on a module
> when a process is using the module. This shows up in the "Used by"
> column of lsmod's output.
>
> The kernel will not let you rmmod a module with a non-zero refcount. You
> cannot get into the situation where you have rmmod'ed the module and a
> reader is still blocking in read()/poll().
However you can still unbind the driver from the device by writing into
driver's sysfs 'unbind' attribute.
See drivers/base/bus.c::driver_unbind().
Thanks.
--
Dmitry
^ permalink raw reply
* Re: [PATCH 1/2] misc: add CARMA DATA-FPGA Access Driver
From: Ira W. Snyder @ 2011-02-09 23:35 UTC (permalink / raw)
To: Dmitry Torokhov; +Cc: linuxppc-dev, linux-kernel
In-Reply-To: <20110209182740.GC23867@core.coreip.homeip.net>
On Wed, Feb 09, 2011 at 10:27:40AM -0800, Dmitry Torokhov wrote:
[ snip stuff I've already fixed in the next version ]
> >
> > The requirement is that the device stay open during reconfiguration.
> > This provides for that. Readers just block for as long as the device is
> > not producing data.
>
> OK, you still need to make sure you do not touch free/used buffer while
> device is disabled. Also, you need to kick readers if you unbind the
> driver, so maybe a new flag priv->exists should be introduced and
> checked.
>
I don't understand what you mean by "kick readers if you unbind the
driver". The kernel automatically increases the refcount on a module
when a process is using the module. This shows up in the "Used by"
column of lsmod's output.
The kernel will not let you rmmod a module with a non-zero refcount. You
cannot get into the situation where you have rmmod'ed the module and a
reader is still blocking in read()/poll().
Thanks for the review. A v6 is coming right up.
Ira
^ permalink raw reply
* [PATCH v3 2/6] powerpc/44x: allow override to hard-coded uart address
From: Dave Kleikamp @ 2011-02-09 23:08 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev, Dave Kleikamp
In-Reply-To: <1297292893-30241-1-git-send-email-shaggy@linux.vnet.ibm.com>
Allow the early debug uart address to be overridden from the kernel
command line.
I would have preferred use the uart's virtual-reg property, but the device
tree hasn't been unflatted yet, and I don't know a reliable way to find it.
Signed-off-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
---
arch/powerpc/kernel/udbg_16550.c | 17 ++++++++++++++---
1 files changed, 14 insertions(+), 3 deletions(-)
diff --git a/arch/powerpc/kernel/udbg_16550.c b/arch/powerpc/kernel/udbg_16550.c
index b4b167b..d36021a 100644
--- a/arch/powerpc/kernel/udbg_16550.c
+++ b/arch/powerpc/kernel/udbg_16550.c
@@ -219,6 +219,19 @@ void udbg_init_pas_realmode(void)
#ifdef CONFIG_PPC_EARLY_DEBUG_44x
#include <platforms/44x/44x.h>
+static unsigned long udbg_44x_comport = PPC44x_EARLY_DEBUG_VIRTADDR;
+
+static int __init early_parse_comport(char *p)
+{
+ if (!p || !(*p))
+ return 0;
+
+ udbg_44x_comport = simple_strtoul(p, 0, 16);
+
+ return 0;
+}
+early_param("uart_addr", early_parse_comport);
+
static void udbg_44x_as1_flush(void)
{
if (udbg_comport) {
@@ -249,9 +262,7 @@ static int udbg_44x_as1_getc(void)
void __init udbg_init_44x_as1(void)
{
- udbg_comport =
- (struct NS16550 __iomem *)PPC44x_EARLY_DEBUG_VIRTADDR;
-
+ udbg_comport = (struct NS16550 __iomem *)udbg_44x_comport;
udbg_putc = udbg_44x_as1_putc;
udbg_flush = udbg_44x_as1_flush;
udbg_getc = udbg_44x_as1_getc;
--
1.7.3.4
^ permalink raw reply related
* [PATCH v3 5/6] powerpc/44x: boot wrapper: allow kernel to load into non-zero address
From: Dave Kleikamp @ 2011-02-09 23:08 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev, Dave Kleikamp
In-Reply-To: <1297292893-30241-1-git-send-email-shaggy@linux.vnet.ibm.com>
For AMP, different kernel instances load into separate memory regions.
Read the start of memory from the device tree and limit the memory to what's
specified in the device tree.
Signed-off-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
---
arch/powerpc/boot/treeboot-iss4xx.c | 22 +++++++++++++++++++++-
1 files changed, 21 insertions(+), 1 deletions(-)
diff --git a/arch/powerpc/boot/treeboot-iss4xx.c b/arch/powerpc/boot/treeboot-iss4xx.c
index fcc4495..674e3bd 100644
--- a/arch/powerpc/boot/treeboot-iss4xx.c
+++ b/arch/powerpc/boot/treeboot-iss4xx.c
@@ -34,9 +34,28 @@
BSS_STACK(4096);
+static u32 ibm4xx_memstart;
+
static void iss_4xx_fixups(void)
{
- ibm4xx_sdram_fixup_memsize();
+ void *memory;
+ u32 reg[3];
+
+ memory = finddevice("/memory");
+ if (!memory)
+ fatal("Can't find memory node\n");
+ getprop(memory, "reg", reg, sizeof(reg));
+ if (reg[2])
+ /* If the device tree specifies the memory range, use it */
+ ibm4xx_memstart = reg[1];
+ else
+ /* othersize, read it from the SDRAM controller */
+ ibm4xx_sdram_fixup_memsize();
+}
+
+static void *iss_4xx_vmlinux_alloc(unsigned long size)
+{
+ return (void *)ibm4xx_memstart;
}
#define SPRN_PIR 0x11E /* Processor Indentification Register */
@@ -48,6 +67,7 @@ void platform_init(void)
simple_alloc_init(_end, avail_ram, 128, 64);
platform_ops.fixups = iss_4xx_fixups;
+ platform_ops.vmlinux_alloc = iss_4xx_vmlinux_alloc;
platform_ops.exit = ibm44x_dbcr_reset;
pir_reg = mfspr(SPRN_PIR);
fdt_set_boot_cpuid_phys(_dtb_start, pir_reg);
--
1.7.3.4
^ permalink raw reply related
* [PATCH v3 3/6] powerpc/47x: allow kernel to be loaded in higher physical memory
From: Dave Kleikamp @ 2011-02-09 23:08 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev, Dave Kleikamp
In-Reply-To: <1297292893-30241-1-git-send-email-shaggy@linux.vnet.ibm.com>
The 44x code (which is shared by 47x) assumes the available physical memory
begins at 0x00000000. This is not necessarily the case in an AMP
environment.
Support CONFIG_RELOCATABLE for 476 in order to allow the kernel to be
loaded into a higher memory range.
Signed-off-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
---
arch/powerpc/Kconfig | 2 +-
arch/powerpc/configs/44x/iss476-smp_defconfig | 6 ++--
arch/powerpc/kernel/head_44x.S | 42 ++++++++++++++++++++-----
arch/powerpc/mm/44x_mmu.c | 13 ++++++--
4 files changed, 48 insertions(+), 15 deletions(-)
diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 7d69e9b..fa41026 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
@@ -827,7 +827,7 @@ config LOWMEM_CAM_NUM
config RELOCATABLE
bool "Build a relocatable kernel (EXPERIMENTAL)"
- depends on EXPERIMENTAL && ADVANCED_OPTIONS && FLATMEM && FSL_BOOKE
+ depends on EXPERIMENTAL && ADVANCED_OPTIONS && FLATMEM && (FSL_BOOKE || PPC_47x)
help
This builds a kernel image that is capable of running at the
location the kernel is loaded at (some alignment restrictions may
diff --git a/arch/powerpc/configs/44x/iss476-smp_defconfig b/arch/powerpc/configs/44x/iss476-smp_defconfig
index 92f863a..a6eb6ad 100644
--- a/arch/powerpc/configs/44x/iss476-smp_defconfig
+++ b/arch/powerpc/configs/44x/iss476-smp_defconfig
@@ -3,8 +3,8 @@ CONFIG_SMP=y
CONFIG_EXPERIMENTAL=y
CONFIG_SYSVIPC=y
CONFIG_POSIX_MQUEUE=y
+CONFIG_SPARSE_IRQ=y
CONFIG_LOG_BUF_SHIFT=14
-CONFIG_SYSFS_DEPRECATED_V2=y
CONFIG_BLK_DEV_INITRD=y
# CONFIG_CC_OPTIMIZE_FOR_SIZE is not set
CONFIG_EXPERT=y
@@ -21,10 +21,11 @@ CONFIG_ISS4xx=y
CONFIG_HZ_100=y
CONFIG_MATH_EMULATION=y
CONFIG_IRQ_ALL_CPUS=y
-CONFIG_SPARSE_IRQ=y
CONFIG_CMDLINE_BOOL=y
CONFIG_CMDLINE="root=/dev/issblk0"
# CONFIG_PCI is not set
+CONFIG_ADVANCED_OPTIONS=y
+CONFIG_RELOCATABLE=y
CONFIG_NET=y
CONFIG_PACKET=y
CONFIG_UNIX=y
@@ -67,7 +68,6 @@ CONFIG_EXT3_FS=y
# CONFIG_EXT3_DEFAULTS_TO_ORDERED is not set
CONFIG_EXT3_FS_POSIX_ACL=y
CONFIG_EXT3_FS_SECURITY=y
-CONFIG_INOTIFY=y
CONFIG_PROC_KCORE=y
CONFIG_TMPFS=y
CONFIG_CRAMFS=y
diff --git a/arch/powerpc/kernel/head_44x.S b/arch/powerpc/kernel/head_44x.S
index cbb3436..1da9b7e 100644
--- a/arch/powerpc/kernel/head_44x.S
+++ b/arch/powerpc/kernel/head_44x.S
@@ -93,6 +93,30 @@ _ENTRY(_start);
bl early_init
+#ifdef CONFIG_RELOCATABLE
+ /*
+ * r25 will contain RPN/ERPN for the start address of memory
+ *
+ * Add the difference between KERNELBASE and PAGE_OFFSET to the
+ * start of physical memory to get kernstart_addr.
+ */
+ lis r3,kernstart_addr@ha
+ la r3,kernstart_addr@l(r3)
+
+ lis r4,KERNELBASE@h
+ ori r4,r4,KERNELBASE@l
+ lis r5,PAGE_OFFSET@h
+ ori r5,r5,PAGE_OFFSET@l
+ subf r4,r5,r4
+
+ rlwinm r6,r25,0,28,31 /* ERPN */
+ rlwinm r7,r25,0,0,3 /* RPN - assuming 256 MB page size */
+ add r7,r7,r4
+
+ stw r6,0(r3)
+ stw r7,4(r3)
+#endif
+
/*
* Decide what sort of machine this is and initialize the MMU.
*/
@@ -1001,9 +1025,6 @@ clear_utlb_entry:
lis r3,PAGE_OFFSET@h
ori r3,r3,PAGE_OFFSET@l
- /* Kernel is at the base of RAM */
- li r4, 0 /* Load the kernel physical address */
-
/* Load the kernel PID = 0 */
li r0,0
mtspr SPRN_PID,r0
@@ -1013,9 +1034,8 @@ clear_utlb_entry:
clrrwi r3,r3,12 /* Mask off the effective page number */
ori r3,r3,PPC47x_TLB0_VALID | PPC47x_TLB0_256M
- /* Word 1 */
- clrrwi r4,r4,12 /* Mask off the real page number */
- /* ERPN is 0 for first 4GB page */
+ /* Word 1 - use r25. RPN is the same as the original entry */
+
/* Word 2 */
li r5,0
ori r5,r5,PPC47x_TLB2_S_RWX
@@ -1026,7 +1046,7 @@ clear_utlb_entry:
/* We write to way 0 and bolted 0 */
lis r0,0x8800
tlbwe r3,r0,0
- tlbwe r4,r0,1
+ tlbwe r25,r0,1
tlbwe r5,r0,2
/*
@@ -1124,7 +1144,13 @@ head_start_common:
lis r4,interrupt_base@h /* IVPR only uses the high 16-bits */
mtspr SPRN_IVPR,r4
- addis r22,r22,KERNELBASE@h
+ /*
+ * If the kernel was loaded at a non-zero 256 MB page, we need to
+ * mask off the most significant 4 bits to get the relative address
+ * from the start of physical memory
+ */
+ rlwinm r22,r22,0,4,31
+ addis r22,r22,PAGE_OFFSET@h
mtlr r22
isync
blr
diff --git a/arch/powerpc/mm/44x_mmu.c b/arch/powerpc/mm/44x_mmu.c
index 024acab..f60e006 100644
--- a/arch/powerpc/mm/44x_mmu.c
+++ b/arch/powerpc/mm/44x_mmu.c
@@ -186,10 +186,11 @@ void __init MMU_init_hw(void)
unsigned long __init mmu_mapin_ram(unsigned long top)
{
unsigned long addr;
+ unsigned long memstart = memstart_addr & ~(PPC_PIN_SIZE - 1);
/* Pin in enough TLBs to cover any lowmem not covered by the
* initial 256M mapping established in head_44x.S */
- for (addr = PPC_PIN_SIZE; addr < lowmem_end_addr;
+ for (addr = memstart + PPC_PIN_SIZE; addr < lowmem_end_addr;
addr += PPC_PIN_SIZE) {
if (mmu_has_feature(MMU_FTR_TYPE_47x))
ppc47x_pin_tlb(addr + PAGE_OFFSET, addr);
@@ -218,19 +219,25 @@ unsigned long __init mmu_mapin_ram(unsigned long top)
void setup_initial_memory_limit(phys_addr_t first_memblock_base,
phys_addr_t first_memblock_size)
{
+ u64 size;
+
+#ifndef CONFIG_RELOCATABLE
/* We don't currently support the first MEMBLOCK not mapping 0
* physical on those processors
*/
BUG_ON(first_memblock_base != 0);
+#endif
/* 44x has a 256M TLB entry pinned at boot */
- memblock_set_current_limit(min_t(u64, first_memblock_size, PPC_PIN_SIZE));
+ size = (min_t(u64, first_memblock_size, PPC_PIN_SIZE));
+ memblock_set_current_limit(first_memblock_base + size);
}
#ifdef CONFIG_SMP
void __cpuinit mmu_init_secondary(int cpu)
{
unsigned long addr;
+ unsigned long memstart = memstart_addr & ~(PPC_PIN_SIZE - 1);
/* Pin in enough TLBs to cover any lowmem not covered by the
* initial 256M mapping established in head_44x.S
@@ -241,7 +248,7 @@ void __cpuinit mmu_init_secondary(int cpu)
* stack. current (r2) isn't initialized, smp_processor_id()
* will not work, current thread info isn't accessible, ...
*/
- for (addr = PPC_PIN_SIZE; addr < lowmem_end_addr;
+ for (addr = memstart + PPC_PIN_SIZE; addr < lowmem_end_addr;
addr += PPC_PIN_SIZE) {
if (mmu_has_feature(MMU_FTR_TYPE_47x))
ppc47x_pin_tlb(addr + PAGE_OFFSET, addr);
--
1.7.3.4
^ permalink raw reply related
* [PATCH v3 1/6] powerpc: Move udbg_early_init() after early_init_devtree()
From: Dave Kleikamp @ 2011-02-09 23:08 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev, Dave Kleikamp
In-Reply-To: <1297292893-30241-1-git-send-email-shaggy@linux.vnet.ibm.com>
so that it can use information from the device tree.
Signed-off-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
---
arch/powerpc/kernel/setup_32.c | 6 +++---
1 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/arch/powerpc/kernel/setup_32.c b/arch/powerpc/kernel/setup_32.c
index 1d2fbc9..d1ca976 100644
--- a/arch/powerpc/kernel/setup_32.c
+++ b/arch/powerpc/kernel/setup_32.c
@@ -120,12 +120,12 @@ notrace void __init machine_init(unsigned long dt_ptr)
{
lockdep_init();
- /* Enable early debugging if any specified (see udbg.h) */
- udbg_early_init();
-
/* Do some early initialization based on the flat device tree */
early_init_devtree(__va(dt_ptr));
+ /* Enable early debugging if any specified (see udbg.h) */
+ udbg_early_init();
+
probe_machine();
setup_kdump_trampoline();
--
1.7.3.4
^ permalink raw reply related
* [PATCH v3 6/6] powerpc/476: Create a dts files for two 476 AMP instances under ISS
From: Dave Kleikamp @ 2011-02-09 23:08 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev, Dave Kleikamp
In-Reply-To: <1297292893-30241-1-git-send-email-shaggy@linux.vnet.ibm.com>
These are completely independent OS instances, each running on 2 cores.
Signed-off-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
---
arch/powerpc/boot/Makefile | 6 ++-
arch/powerpc/boot/dts/iss476-amp1.dts | 119 ++++++++++++++++++++++++++++++++
arch/powerpc/boot/dts/iss476-amp2.dts | 120 +++++++++++++++++++++++++++++++++
arch/powerpc/boot/wrapper | 7 ++
4 files changed, 251 insertions(+), 1 deletions(-)
create mode 100644 arch/powerpc/boot/dts/iss476-amp1.dts
create mode 100644 arch/powerpc/boot/dts/iss476-amp2.dts
diff --git a/arch/powerpc/boot/Makefile b/arch/powerpc/boot/Makefile
index 8917816..99dbc39 100644
--- a/arch/powerpc/boot/Makefile
+++ b/arch/powerpc/boot/Makefile
@@ -45,6 +45,8 @@ $(obj)/cuboot-katmai.o: BOOTCFLAGS += -mcpu=405
$(obj)/cuboot-acadia.o: BOOTCFLAGS += -mcpu=405
$(obj)/treeboot-walnut.o: BOOTCFLAGS += -mcpu=405
$(obj)/treeboot-iss4xx.o: BOOTCFLAGS += -mcpu=405
+$(obj)/treeboot-iss476-amp1.o: BOOTCFLAGS += -mcpu=405
+$(obj)/treeboot-iss476-amp2.o: BOOTCFLAGS += -mcpu=405
$(obj)/virtex405-head.o: BOOTAFLAGS += -mcpu=405
@@ -208,7 +210,9 @@ image-$(CONFIG_KATMAI) += cuImage.katmai
image-$(CONFIG_WARP) += cuImage.warp
image-$(CONFIG_YOSEMITE) += cuImage.yosemite
image-$(CONFIG_ISS4xx) += treeImage.iss4xx \
- treeImage.iss4xx-mpic
+ treeImage.iss4xx-mpic \
+ treeImage.iss476-amp1 \
+ treeImage.iss476-amp2
# Board ports in arch/powerpc/platform/8xx/Kconfig
image-$(CONFIG_MPC86XADS) += cuImage.mpc866ads
diff --git a/arch/powerpc/boot/dts/iss476-amp1.dts b/arch/powerpc/boot/dts/iss476-amp1.dts
new file mode 100644
index 0000000..1fa882c
--- /dev/null
+++ b/arch/powerpc/boot/dts/iss476-amp1.dts
@@ -0,0 +1,119 @@
+/*
+ * Device Tree Source for IBM Embedded PPC 476 Platform
+ *
+ * Copyright 2010 Torez Smith, IBM Corporation.
+ *
+ * Based on earlier code:
+ * Copyright (c) 2006, 2007 IBM Corp.
+ * Josh Boyer <jwboyer@linux.vnet.ibm.com>, David Gibson <dwg@au1.ibm.com>
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2. This program is licensed "as is" without
+ * any warranty of any kind, whether express or implied.
+ */
+
+/dts-v1/;
+
+/ {
+ #address-cells = <2>;
+ #size-cells = <1>;
+ model = "ibm,iss-4xx";
+ compatible = "ibm,iss-4xx";
+ cooperative-partition;
+ dcr-parent = <&{/cpus/cpu@0}>;
+
+ aliases {
+ serial0 = &UART0;
+ };
+
+ cpus {
+ #address-cells = <1>;
+ #size-cells = <0>;
+
+ cpu@0 {
+ device_type = "cpu";
+ model = "PowerPC,4xx";
+ reg = <0>;
+ clock-frequency = <100000000>; // 100Mhz :-)
+ timebase-frequency = <100000000>;
+ i-cache-line-size = <32>;
+ d-cache-line-size = <32>;
+ i-cache-size = <32768>;
+ d-cache-size = <32768>;
+ dcr-controller;
+ dcr-access-method = "native";
+ status = "okay";
+ };
+ cpu@1 {
+ device_type = "cpu";
+ model = "PowerPC,4xx"; // real CPU changed in sim
+ reg = <1>;
+ clock-frequency = <100000000>; // 100Mhz :-)
+ timebase-frequency = <100000000>;
+ i-cache-line-size = <32>;
+ d-cache-line-size = <32>;
+ i-cache-size = <32768>;
+ d-cache-size = <32768>;
+ dcr-controller;
+ dcr-access-method = "native";
+ status = "disabled";
+ enable-method = "spin-table";
+ cpu-release-addr = <0 0x01f00100>;
+ };
+ };
+
+ memory {
+ device_type = "memory";
+ reg = <0x00000000 0x00000000 0x02000000>;
+
+ };
+
+ MPIC: interrupt-controller {
+ compatible = "chrp,open-pic";
+ interrupt-controller;
+ dcr-reg = <0xffc00000 0x00030000>;
+ #address-cells = <0>;
+ #size-cells = <0>;
+ #interrupt-cells = <2>;
+ protected-sources = <1>; /* uart */
+
+ };
+
+ plb {
+ compatible = "ibm,plb-4xx", "ibm,plb4"; /* Could be PLB6, doesn't matter */
+ #address-cells = <2>;
+ #size-cells = <1>;
+ ranges;
+ clock-frequency = <0>; // Filled in by zImage
+
+ POB0: opb {
+ compatible = "ibm,opb-4xx", "ibm,opb";
+ #address-cells = <1>;
+ #size-cells = <1>;
+ /* Wish there was a nicer way of specifying a full 32-bit
+ range */
+ ranges = <0x00000000 0x00000001 0x00000000 0x80000000
+ 0x80000000 0x00000001 0x80000000 0x80000000>;
+ clock-frequency = <0>; // Filled in by zImage
+ UART0: serial@40000200 {
+ device_type = "serial";
+ compatible = "ns16550a";
+ reg = <0x40000200 0x00000008>;
+ virtual-reg = <0xe0000200>;
+ clock-frequency = <11059200>;
+ current-speed = <115200>;
+ interrupt-parent = <&MPIC>;
+ interrupts = <0x0 0x2>;
+ };
+ };
+ };
+
+ nvrtc {
+ compatible = "ds1743-nvram", "ds1743", "rtc-ds1743";
+ reg = <0 0xEF703000 0x2000>;
+ };
+
+ chosen {
+ linux,stdout-path = &UART0;
+ };
+};
diff --git a/arch/powerpc/boot/dts/iss476-amp2.dts b/arch/powerpc/boot/dts/iss476-amp2.dts
new file mode 100644
index 0000000..09b4c57
--- /dev/null
+++ b/arch/powerpc/boot/dts/iss476-amp2.dts
@@ -0,0 +1,120 @@
+/*
+ * Device Tree Source for IBM Embedded PPC 476 Platform
+ *
+ * Copyright 2010 Torez Smith, IBM Corporation.
+ *
+ * Based on earlier code:
+ * Copyright (c) 2006, 2007 IBM Corp.
+ * Josh Boyer <jwboyer@linux.vnet.ibm.com>, David Gibson <dwg@au1.ibm.com>
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2. This program is licensed "as is" without
+ * any warranty of any kind, whether express or implied.
+ */
+
+/dts-v1/;
+
+/ {
+ #address-cells = <2>;
+ #size-cells = <1>;
+ model = "ibm,iss-4xx";
+ compatible = "ibm,iss-4xx";
+ cooperative-partition;
+ dcr-parent = <&{/cpus/cpu@2}>;
+
+ aliases {
+ serial0 = &UART0;
+ };
+
+ cpus {
+ #address-cells = <1>;
+ #size-cells = <0>;
+
+ cpu@2 {
+ device_type = "cpu";
+ model = "PowerPC,4xx";
+ reg = <2>;
+ clock-frequency = <100000000>; // 100Mhz :-)
+ timebase-frequency = <100000000>;
+ i-cache-line-size = <32>;
+ d-cache-line-size = <32>;
+ i-cache-size = <32768>;
+ d-cache-size = <32768>;
+ dcr-controller;
+ dcr-access-method = "native";
+ status = "okay";
+ };
+ cpu@3 {
+ device_type = "cpu";
+ model = "PowerPC,4xx"; // real CPU changed in sim
+ reg = <3>;
+ clock-frequency = <100000000>; // 100Mhz :-)
+ timebase-frequency = <100000000>;
+ i-cache-line-size = <32>;
+ d-cache-line-size = <32>;
+ i-cache-size = <32768>;
+ d-cache-size = <32768>;
+ dcr-controller;
+ dcr-access-method = "native";
+ status = "disabled";
+ enable-method = "spin-table";
+ cpu-release-addr = <0 0x11f00300>;
+ };
+ };
+
+ memory {
+ device_type = "memory";
+ reg = < 0x0 0x10000000 0x02000000 >;
+
+ };
+
+ MPIC: interrupt-controller {
+ compatible = "chrp,open-pic";
+ interrupt-controller;
+ dcr-reg = <0xffc00000 0x00030000>;
+ #address-cells = <0>;
+ #size-cells = <0>;
+ #interrupt-cells = <2>;
+ protected-sources = <0>; /* uart */
+
+ };
+
+ plb {
+ compatible = "ibm,plb-4xx", "ibm,plb4"; /* Could be PLB6, doesn't matter */
+ #address-cells = <2>;
+ #size-cells = <1>;
+ ranges;
+ clock-frequency = <0>; // Filled in by zImage
+
+ POB0: opb {
+ compatible = "ibm,opb-4xx", "ibm,opb";
+ #address-cells = <1>;
+ #size-cells = <1>;
+ /* Wish there was a nicer way of specifying a full 32-bit
+ range */
+ ranges = <0x00000000 0x00000001 0x00000000 0x80000000
+ 0x80000000 0x00000001 0x80000000 0x80000000>;
+ clock-frequency = <0>; // Filled in by zImage
+ UART0: serial@40001200 {
+ device_type = "serial";
+ compatible = "ns16550a";
+ reg = <0x40001200 0x00000008>;
+ virtual-reg = <0xe0001200>;
+ clock-frequency = <11059200>;
+ current-speed = <115200>;
+ interrupt-parent = <&MPIC>;
+ interrupts = <0x1 0x2>;
+ };
+ };
+ };
+
+ nvrtc {
+ compatible = "ds1743-nvram", "ds1743", "rtc-ds1743";
+ reg = <0 0xEF703000 0x2000>;
+ };
+
+ chosen {
+ bootargs = "uart_addr=0xf0001200";
+ linux,stdout-path = &UART0;
+ };
+};
diff --git a/arch/powerpc/boot/wrapper b/arch/powerpc/boot/wrapper
index cb97e75..8e6ca36 100755
--- a/arch/powerpc/boot/wrapper
+++ b/arch/powerpc/boot/wrapper
@@ -244,6 +244,13 @@ gamecube|wii)
treeboot-iss4xx-mpic)
platformo="$object/treeboot-iss4xx.o"
;;
+treeboot-iss476-amp1)
+ platformo="$object/treeboot-iss4xx.o"
+ ;;
+treeboot-iss476-amp2)
+ platformo="$object/treeboot-iss4xx.o"
+ link_address='0x10400000'
+ ;;
esac
vmz="$tmpdir/`basename \"$kernel\"`.$ext"
--
1.7.3.4
^ permalink raw reply related
* [PATCH v3 4/6] powerpc/44x: don't use tlbivax on AMP systems
From: Dave Kleikamp @ 2011-02-09 23:08 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev, Dave Kleikamp
In-Reply-To: <1297292893-30241-1-git-send-email-shaggy@linux.vnet.ibm.com>
Since other OS's may be running on the other cores don't use tlbivax
Signed-off-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
---
arch/powerpc/include/asm/mmu.h | 2 +-
arch/powerpc/kernel/setup_32.c | 2 ++
arch/powerpc/mm/tlb_nohash.c | 23 ++++++++++++++++++++++-
3 files changed, 25 insertions(+), 2 deletions(-)
diff --git a/arch/powerpc/include/asm/mmu.h b/arch/powerpc/include/asm/mmu.h
index bb40a06..f3a7c65 100644
--- a/arch/powerpc/include/asm/mmu.h
+++ b/arch/powerpc/include/asm/mmu.h
@@ -80,7 +80,7 @@ static inline int mmu_has_feature(unsigned long feature)
extern unsigned int __start___mmu_ftr_fixup, __stop___mmu_ftr_fixup;
-/* MMU initialization (64-bit only fo now) */
+/* MMU initialization */
extern void early_init_mmu(void);
extern void early_init_mmu_secondary(void);
diff --git a/arch/powerpc/kernel/setup_32.c b/arch/powerpc/kernel/setup_32.c
index d1ca976..e50ead7 100644
--- a/arch/powerpc/kernel/setup_32.c
+++ b/arch/powerpc/kernel/setup_32.c
@@ -126,6 +126,8 @@ notrace void __init machine_init(unsigned long dt_ptr)
/* Enable early debugging if any specified (see udbg.h) */
udbg_early_init();
+ early_init_mmu();
+
probe_machine();
setup_kdump_trampoline();
diff --git a/arch/powerpc/mm/tlb_nohash.c b/arch/powerpc/mm/tlb_nohash.c
index 2a030d8..5f753b8 100644
--- a/arch/powerpc/mm/tlb_nohash.c
+++ b/arch/powerpc/mm/tlb_nohash.c
@@ -35,6 +35,7 @@
#include <linux/preempt.h>
#include <linux/spinlock.h>
#include <linux/memblock.h>
+#include <linux/of_fdt.h>
#include <asm/tlbflush.h>
#include <asm/tlb.h>
@@ -153,6 +154,8 @@ EXPORT_SYMBOL(local_flush_tlb_page);
*/
#ifdef CONFIG_SMP
+static int amp;
+
static DEFINE_RAW_SPINLOCK(tlbivax_lock);
static int mm_is_core_local(struct mm_struct *mm)
@@ -232,7 +235,7 @@ void __flush_tlb_page(struct mm_struct *mm, unsigned long vmaddr,
cpu_mask = mm_cpumask(mm);
if (!mm_is_core_local(mm)) {
/* If broadcast tlbivax is supported, use it */
- if (mmu_has_feature(MMU_FTR_USE_TLBIVAX_BCAST)) {
+ if (!amp && mmu_has_feature(MMU_FTR_USE_TLBIVAX_BCAST)) {
int lock = mmu_has_feature(MMU_FTR_LOCK_BCAST_INVAL);
if (lock)
raw_spin_lock(&tlbivax_lock);
@@ -266,6 +269,17 @@ EXPORT_SYMBOL(flush_tlb_page);
#endif /* CONFIG_SMP */
+#ifdef CONFIG_PPC_47x
+void __init early_init_mmu_47x(void)
+{
+#ifdef CONFIG_SMP
+ unsigned long root = of_get_flat_dt_root();
+ if (of_get_flat_dt_prop(root, "cooperative-partition", NULL))
+ amp = 1;
+#endif /* CONFIG_SMP */
+}
+#endif /* CONFIG_PPC_47x */
+
/*
* Flush kernel TLB entries in the given range
*/
@@ -587,4 +601,11 @@ void setup_initial_memory_limit(phys_addr_t first_memblock_base,
/* Finally limit subsequent allocations */
memblock_set_current_limit(first_memblock_base + ppc64_rma_size);
}
+#else /* ! CONFIG_PPC64 */
+void __init early_init_mmu(void)
+{
+#ifdef CONFIG_PPC_47x
+ early_init_mmu_47x();
+#endif
+}
#endif /* CONFIG_PPC64 */
--
1.7.3.4
^ permalink raw reply related
* [PATCH v3 0/6] powerpc: AMP support for 47x
From: Dave Kleikamp @ 2011-02-09 23:08 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev, Dave Kleikamp
These patches add Asynchonous MultiProcessing support for the 47x chipset.
This allows independent OS instances to run on separate cores.
v3:
1. switched from using compatible string to boolean "cooperative-partition"
2. fixed missing type in boot wrapper
3. fixed check for unspecified memory range in boot wrapper
4. Cleaned up the dts files
v2:
1. Replace ugly hack in boot wrapper with generic solution
Dave Kleikamp (6):
powerpc: Move udbg_early_init() after early_init_devtree()
powerpc/44x: allow override to hard-coded uart address
powerpc/47x: allow kernel to be loaded in higher physical memory
powerpc/44x: don't use tlbivax on AMP systems
powerpc/44x: boot wrapper: allow kernel to load into non-zero address
powerpc/476: Create a dts files for two 476 AMP instances under ISS
arch/powerpc/Kconfig | 2 +-
arch/powerpc/boot/Makefile | 6 +-
arch/powerpc/boot/dts/iss476-amp1.dts | 119 ++++++++++++++++++++++++
arch/powerpc/boot/dts/iss476-amp2.dts | 120 +++++++++++++++++++++++++
arch/powerpc/boot/treeboot-iss4xx.c | 22 +++++-
arch/powerpc/boot/wrapper | 7 ++
arch/powerpc/configs/44x/iss476-smp_defconfig | 6 +-
arch/powerpc/include/asm/mmu.h | 2 +-
arch/powerpc/kernel/head_44x.S | 42 +++++++--
arch/powerpc/kernel/setup_32.c | 6 +-
arch/powerpc/kernel/udbg_16550.c | 17 +++-
arch/powerpc/mm/44x_mmu.c | 13 ++-
arch/powerpc/mm/tlb_nohash.c | 23 +++++-
13 files changed, 361 insertions(+), 24 deletions(-)
create mode 100644 arch/powerpc/boot/dts/iss476-amp1.dts
create mode 100644 arch/powerpc/boot/dts/iss476-amp2.dts
--
1.7.3.4
^ permalink raw reply
* Re: [PATCH V2 6/6] powerpc/476: Create a dts files for two 476 AMP instances under ISS
From: Dave Kleikamp @ 2011-02-09 23:03 UTC (permalink / raw)
To: David Gibson; +Cc: linuxppc-dev
In-Reply-To: <20110202024315.GB3032@yookeroo>
On Wed, 2011-02-02 at 13:43 +1100, David Gibson wrote:
> On Tue, Feb 01, 2011 at 12:48:46PM -0600, Dave Kleikamp wrote:
> > These are completely independent OS instances, each running on 2
> > cores.
>
> [snip]
> > +/memreserve/ 0x01f00000 0x00100000;
>
> A comment describing what this reserved section is for would be good.
I with I knew what it was for. I've blindly carried it along for a
while. Removing it doesn't appear to do any harm. Ben, any idea why
this was ever in here?
> > +/ {
> > + #address-cells = <2>;
> > + #size-cells = <1>;
> > + model = "ibm,iss-4xx";
> > + compatible = "ibm,iss-4xx", "ibm,47x-AMP";
> > + dcr-parent = <&{/cpus/cpu@0}>;
> > +
> > + aliases {
> > + serial0 = &UART0;
> > + };
> > +
> > + cpus {
> > + #address-cells = <1>;
> > + #size-cells = <0>;
> > +
> > + cpu@0 {
> > + device_type = "cpu";
> > + model = "PowerPC,4xx"; // real CPU changed in sim
>
> If the comment is true, then it's probably simpler to just omit the
> model property. I'm pretty sure nothing will look at it.
It doesn't appear to be true. Another bit I've been carrying along
without checking it. Removing the comment.
>
> > + reg = <0>;
> > + clock-frequency = <100000000>; // 100Mhz :-)
> > + timebase-frequency = <100000000>;> + i-cache-line-size = <32>;
> > + d-cache-line-size = <32>;
> > + i-cache-size = <32768>;
> > + d-cache-size = <32768>;
> > + dcr-controller;
> > + dcr-access-method = "native";
> > + status = "ok";
>
> Should be "okay" rather than "ok".
okay :-)
>
> [snip]
> > + UART0: serial@40000200 {
> > + device_type = "serial";
> > + compatible = "ns16550a";
> > + reg = <0x40000200 0x00000008>;
> > + virtual-reg = <0xe0000200>;
> > + clock-frequency = <11059200>;
> > + current-speed = <115200>;
> > + interrupt-parent = <&MPIC>;
> > + interrupts = <0x0 0x2>;
> > + };
> > + };
> > + };
> > +
> > + nvrtc {
> > + compatible = "ds1743-nvram", "ds1743", "rtc-ds1743";
> > + reg = <0 0xEF703000 0x2000>;
> > + };
> > +
> > + chosen {
> > + linux,stdout-path = "/plb/opb/serial@40000200";
>
> You can use a string reference here:
> linux,stdout-path = &UART0;
no problem
Shaggy
--
Dave Kleikamp
IBM Linux Technology Center
^ permalink raw reply
* Re: [PATCH V2 3/6] powerpc/47x: allow kernel to be loaded in higher physical memory
From: Dave Kleikamp @ 2011-02-09 22:59 UTC (permalink / raw)
To: Kumar Gala; +Cc: linuxppc-dev
In-Reply-To: <7637207D-0D0C-4920-9E8D-B175711C1164@kernel.crashing.org>
On Wed, 2011-02-02 at 01:45 -0600, Kumar Gala wrote:
> On Feb 1, 2011, at 12:48 PM, Dave Kleikamp wrote:
>
> > Signed-off-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
> > Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> > Cc: Josh Boyer <jwboyer@linux.vnet.ibm.com>
> > Cc: linuxppc-dev@lists.ozlabs.org
> > ---
> > arch/powerpc/Kconfig | 2 +-
> > arch/powerpc/configs/44x/iss476-smp_defconfig | 6 ++--
> > arch/powerpc/kernel/head_44x.S | 42 ++++++++++++++++++++-----
> > arch/powerpc/mm/44x_mmu.c | 13 ++++++--
> > 4 files changed, 48 insertions(+), 15 deletions(-)
>
> Would be nice to expand in the commit message on what "higher physical" means.
right. should be clear in the next set
--
Dave Kleikamp
IBM Linux Technology Center
^ permalink raw reply
* Re: [PATCH 2/6] nvram: Capture oops/panic reports in ibm, oops-log partition
From: Jim Keniston @ 2011-02-09 23:00 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev
In-Reply-To: <1297054887.14982.55.camel@pasglop>
On Mon, 2011-02-07 at 16:01 +1100, Benjamin Herrenschmidt wrote:
> On Sat, 2010-11-13 at 20:15 -0800, Jim Keniston wrote:
> > Create the ibm,oops-log NVRAM partition, and capture the end of the printk
> > buffer in it when there's an oops or panic. If we can't create the
> > ibm,oops-log partition, capture the oops/panic report in ibm,rtas-log.
I've appended an updated patch that addresses your comments and applies
to 2.6.38-rc4.
>
> Won't the parsing tools choke on the oops log in the RTAS log
> partition ?
I don't think that that will be a problem, because pSeries_log_error()
doesn't log errors that aren't of type ERR_TYPE_RTAS_LOG.
However, I did add code to ensure that even if ibm,rtas-log is used for
both RTAS errors and oops reports, an oops report won't overwrite a
recently captured RTAS event that hasn't yet been read into the
filesystem (e.g., by rtas_errd).
>
> Also, maybe remove the "ibm," prefix on the nvram partition, make it
> "lnx," instead.
Done.
>
> Then, if you move the rest of the code out of pseries, you can move that
> out too, just make pseries platform code call the init code for it, that
> way other platforms can re-use that as-is.
>
> But that later part can wait a separate patch series,
Agreed.
> just make sure you
> change the partition name now.
Still done.
>
> Cheers,
> Ben.
Here's my revised patch.
Jim
=====
Capture oops/panic reports in lnx,oops-log partition
Create the lnx,oops-log NVRAM partition, and capture the end of the printk
buffer in it when there's an oops or panic. If we can't create the
lnx,oops-log partition, capture the oops/panic report in ibm,rtas-log.
Signed-off-by: Jim Keniston <jkenisto@us.ibm.com>
---
arch/powerpc/platforms/pseries/nvram.c | 118 +++++++++++++++++++++++++++++++-
1 files changed, 115 insertions(+), 3 deletions(-)
diff --git a/arch/powerpc/platforms/pseries/nvram.c b/arch/powerpc/platforms/pseries/nvram.c
index befb41b..419707b 100644
--- a/arch/powerpc/platforms/pseries/nvram.c
+++ b/arch/powerpc/platforms/pseries/nvram.c
@@ -16,6 +16,8 @@
#include <linux/errno.h>
#include <linux/init.h>
#include <linux/spinlock.h>
+#include <linux/slab.h>
+#include <linux/kmsg_dump.h>
#include <asm/uaccess.h>
#include <asm/nvram.h>
#include <asm/rtas.h>
@@ -39,7 +41,7 @@ struct nvram_os_partition {
const char *name;
int req_size; /* desired size, in bytes */
int min_size; /* minimum acceptable size (0 means req_size) */
- long size; /* size of data portion of partition */
+ long size; /* size of data portion (excluding err_log_info) */
long index; /* offset of data portion of partition */
};
@@ -50,11 +52,35 @@ static struct nvram_os_partition rtas_log_partition = {
.index = -1
};
+static struct nvram_os_partition oops_log_partition = {
+ .name = "lnx,oops-log",
+ .req_size = 4000,
+ .min_size = 2000,
+ .index = -1
+};
+
static const char *pseries_nvram_os_partitions[] = {
"ibm,rtas-log",
+ "lnx,oops-log",
NULL
};
+static void oops_to_nvram(struct kmsg_dumper *dumper,
+ enum kmsg_dump_reason reason,
+ const char *old_msgs, unsigned long old_len,
+ const char *new_msgs, unsigned long new_len);
+
+static struct kmsg_dumper nvram_kmsg_dumper = {
+ .dump = oops_to_nvram
+};
+
+/* See clobbering_unread_rtas_event() */
+#define NVRAM_RTAS_READ_TIMEOUT 5 /* seconds */
+static unsigned long last_unread_rtas_event; /* timestamp */
+
+/* We preallocate oops_buf during init to avoid kmalloc during oops/panic. */
+static char *oops_buf;
+
static ssize_t pSeries_nvram_read(char *buf, size_t count, loff_t *index)
{
unsigned int i;
@@ -214,8 +240,11 @@ int nvram_write_os_partition(struct nvram_os_partition *part, char * buff,
int nvram_write_error_log(char * buff, int length,
unsigned int err_type, unsigned int error_log_cnt)
{
- return nvram_write_os_partition(&rtas_log_partition, buff, length,
+ int rc = nvram_write_os_partition(&rtas_log_partition, buff, length,
err_type, error_log_cnt);
+ if (!rc)
+ last_unread_rtas_event = get_seconds();
+ return rc;
}
/* nvram_read_error_log
@@ -274,6 +303,7 @@ int nvram_clear_error_log(void)
printk(KERN_ERR "nvram_clear_error_log: Failed nvram_write (%d)\n", rc);
return rc;
}
+ last_unread_rtas_event = 0;
return 0;
}
@@ -342,9 +372,35 @@ static int __init pseries_nvram_init_os_partition(struct nvram_os_partition
return 0;
}
+static void __init nvram_init_oops_partition(int rtas_partition_exists)
+{
+ int rc;
+
+ rc = pseries_nvram_init_os_partition(&oops_log_partition);
+ if (rc != 0) {
+ if (!rtas_partition_exists)
+ return;
+ pr_notice("nvram: Using %s partition to log both"
+ " RTAS errors and oops/panic reports\n",
+ rtas_log_partition.name);
+ memcpy(&oops_log_partition, &rtas_log_partition,
+ sizeof(rtas_log_partition));
+ }
+ oops_buf = kmalloc(oops_log_partition.size, GFP_KERNEL);
+ rc = kmsg_dump_register(&nvram_kmsg_dumper);
+ if (rc != 0) {
+ pr_err("nvram: kmsg_dump_register() failed; returned %d\n", rc);
+ kfree(oops_buf);
+ return;
+ }
+}
+
static int __init pseries_nvram_init_log_partitions(void)
{
- (void) pseries_nvram_init_os_partition(&rtas_log_partition);
+ int rc;
+
+ rc = pseries_nvram_init_os_partition(&rtas_log_partition);
+ nvram_init_oops_partition(rc == 0);
return 0;
}
machine_arch_initcall(pseries, pseries_nvram_init_log_partitions);
@@ -378,3 +434,59 @@ int __init pSeries_nvram_init(void)
return 0;
}
+
+/*
+ * Try to capture the last capture_len bytes of the printk buffer. Return
+ * the amount actually captured.
+ */
+static size_t capture_last_msgs(const char *old_msgs, size_t old_len,
+ const char *new_msgs, size_t new_len,
+ char *captured, size_t capture_len)
+{
+ if (new_len >= capture_len) {
+ memcpy(captured, new_msgs + (new_len - capture_len),
+ capture_len);
+ return capture_len;
+ } else {
+ /* Grab the end of old_msgs. */
+ size_t old_tail_len = min(old_len, capture_len - new_len);
+ memcpy(captured, old_msgs + (old_len - old_tail_len),
+ old_tail_len);
+ memcpy(captured + old_tail_len, new_msgs, new_len);
+ return old_tail_len + new_len;
+ }
+}
+
+/*
+ * Are we using the ibm,rtas-log for oops/panic reports? And if so,
+ * would logging this oops/panic overwrite an RTAS event that rtas_errd
+ * hasn't had a chance to read and process? Return 1 if so, else 0.
+ *
+ * We assume that if rtas_errd hasn't read the RTAS event in
+ * NVRAM_RTAS_READ_TIMEOUT seconds, it's probably not going to.
+ */
+static int clobbering_unread_rtas_event(void)
+{
+ return (oops_log_partition.index == rtas_log_partition.index
+ && last_unread_rtas_event
+ && get_seconds() - last_unread_rtas_event <=
+ NVRAM_RTAS_READ_TIMEOUT);
+}
+
+/* our kmsg_dump callback */
+static void oops_to_nvram(struct kmsg_dumper *dumper,
+ enum kmsg_dump_reason reason,
+ const char *old_msgs, unsigned long old_len,
+ const char *new_msgs, unsigned long new_len)
+{
+ static unsigned int oops_count = 0;
+ size_t text_len;
+
+ if (clobbering_unread_rtas_event())
+ return;
+
+ text_len = capture_last_msgs(old_msgs, old_len, new_msgs, new_len,
+ oops_buf, oops_log_partition.size);
+ (void) nvram_write_os_partition(&oops_log_partition, oops_buf,
+ (int) text_len, ERR_TYPE_KERNEL_PANIC, ++oops_count);
+}
^ permalink raw reply related
* Re: [PATCH 1/6] nvram: Generalize code for OS partitions in NVRAM
From: Jim Keniston @ 2011-02-09 22:43 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev
In-Reply-To: <1297054622.14982.51.camel@pasglop>
On Mon, 2011-02-07 at 15:57 +1100, Benjamin Herrenschmidt wrote:
> On Sat, 2010-11-13 at 20:15 -0800, Jim Keniston wrote:
> > Adapt the functions used to create and write to the RTAS-log partition
> > to work with any OS-type partition.
> >
> > Signed-off-by: Jim Keniston <jkenisto@us.ibm.com>
> > ---
>
> Overall pretty good (sorry for taking that long to review, I've been
> swamped down). Just a handful of minor nits:
I've appended an updated patch, which addresses your comments and also
applies to 2.6.38-rc4, rather than to your kernel as it stood on Nov.
13. See below for responses to comments.
>
> > +/*
> > + * Per the criteria passed via nvram_remove_partition(), should this
> > + * partition be removed? 1=remove, 0=keep
> > + */
> > +static int nvram_condemn_partition(struct nvram_partition *part,
> > + const char *name, int sig, const char *exceptions[])
>
> As "fun" as this name is, it didn't help me understand what the function
> was about until I read the code for the next one :-)
>
> What about instead something like nvram_can_remove_partition() or
> something a bit more explicit like that ?
Done.
...
> .../...
>
> > +static const char *valid_os_partitions[] = {
> > + "ibm,rtas-log",
> > + NULL
> > +};
>
> Can you pick a name that will be less confusing in the global
> symbol table ? Something like "pseries_nvram_os_partitions"
Done.
> or whatever shorter you can come up with which doesn't suck ?
>
> Also, should we move that "os partition" core out of pseries ?
>
> Looks like it will be useful for a few other platforms, I'd like
> to move that to a more generically useful location in arch/powerpc
> but that's not a blocker for this series but something to do next
> maybe ?
I haven't tried to do that reorganization.
>
> In that case "struct os_partition" should also find itself a better
> name, maybe struct nvram_os_partition ?
Done.
...
> > if (rc <= 0) {
> > - printk(KERN_ERR "nvram_write_error_log: Failed nvram_write (%d)\n", rc);
> > + printk(KERN_ERR "nvram_write_os_partition: Failed nvram_write (%d)\n", rc);
> > return rc;
> > }
> >
> > return 0;
> > }
>
> While at it, turn these into pr_err and use __FUNCTION__ or __func__
Done. However, I didn't convert any printks in functions that I wasn't
otherwise changing.
>
> > +int nvram_write_error_log(char * buff, int length,
> > + unsigned int err_type, unsigned int error_log_cnt)
> > +{
> > + return nvram_write_os_partition(&rtas_log_partition, buff, length,
> > + err_type, error_log_cnt);
> > +}
>
> That's a candidate for a static inline in a .h
Maybe, but in my next patch this function adjusts a static variable, so
I figure that putting it in a header would be messy.
Here's my revised patch.
Jim
=====
Generalize code for OS partitions in NVRAM
Adapt the functions used to create and write to the RTAS-log partition
to work with any OS-type partition.
Signed-off-by: Jim Keniston <jkenisto@us.ibm.com>
---
arch/powerpc/include/asm/nvram.h | 3 -
arch/powerpc/kernel/nvram_64.c | 31 ++++++-
arch/powerpc/platforms/pseries/nvram.c | 143 +++++++++++++++++++-------------
3 files changed, 113 insertions(+), 64 deletions(-)
diff --git a/arch/powerpc/include/asm/nvram.h b/arch/powerpc/include/asm/nvram.h
index 92efe67..9d1aafe 100644
--- a/arch/powerpc/include/asm/nvram.h
+++ b/arch/powerpc/include/asm/nvram.h
@@ -51,7 +51,8 @@ static inline int mmio_nvram_init(void)
extern int __init nvram_scan_partitions(void);
extern loff_t nvram_create_partition(const char *name, int sig,
int req_size, int min_size);
-extern int nvram_remove_partition(const char *name, int sig);
+extern int nvram_remove_partition(const char *name, int sig,
+ const char *exceptions[]);
extern int nvram_get_partition_size(loff_t data_index);
extern loff_t nvram_find_partition(const char *name, int sig, int *out_size);
diff --git a/arch/powerpc/kernel/nvram_64.c b/arch/powerpc/kernel/nvram_64.c
index bb12b32..bec1e93 100644
--- a/arch/powerpc/kernel/nvram_64.c
+++ b/arch/powerpc/kernel/nvram_64.c
@@ -237,22 +237,45 @@ static unsigned char __init nvram_checksum(struct nvram_header *p)
return c_sum;
}
+/*
+ * Per the criteria passed via nvram_remove_partition(), should this
+ * partition be removed? 1=remove, 0=keep
+ */
+static int nvram_can_remove_partition(struct nvram_partition *part,
+ const char *name, int sig, const char *exceptions[])
+{
+ if (part->header.signature != sig)
+ return 0;
+ if (name) {
+ if (strncmp(name, part->header.name, 12))
+ return 0;
+ } else if (exceptions) {
+ const char **except;
+ for (except = exceptions; *except; except++) {
+ if (!strncmp(*except, part->header.name, 12))
+ return 0;
+ }
+ }
+ return 1;
+}
+
/**
* nvram_remove_partition - Remove one or more partitions in nvram
* @name: name of the partition to remove, or NULL for a
* signature only match
* @sig: signature of the partition(s) to remove
+ * @exceptions: When removing all partitions with a matching signature,
+ * leave these alone.
*/
-int __init nvram_remove_partition(const char *name, int sig)
+int __init nvram_remove_partition(const char *name, int sig,
+ const char *exceptions[])
{
struct nvram_partition *part, *prev, *tmp;
int rc;
list_for_each_entry(part, &nvram_partitions, partition) {
- if (part->header.signature != sig)
- continue;
- if (name && strncmp(name, part->header.name, 12))
+ if (!nvram_can_remove_partition(part, name, sig, exceptions))
continue;
/* Make partition a free partition */
diff --git a/arch/powerpc/platforms/pseries/nvram.c b/arch/powerpc/platforms/pseries/nvram.c
index 7e828ba..befb41b 100644
--- a/arch/powerpc/platforms/pseries/nvram.c
+++ b/arch/powerpc/platforms/pseries/nvram.c
@@ -30,17 +30,30 @@ static int nvram_fetch, nvram_store;
static char nvram_buf[NVRW_CNT]; /* assume this is in the first 4GB */
static DEFINE_SPINLOCK(nvram_lock);
-static long nvram_error_log_index = -1;
-static long nvram_error_log_size = 0;
-
struct err_log_info {
int error_type;
unsigned int seq_num;
};
-#define NVRAM_MAX_REQ 2079
-#define NVRAM_MIN_REQ 1055
-#define NVRAM_LOG_PART_NAME "ibm,rtas-log"
+struct nvram_os_partition {
+ const char *name;
+ int req_size; /* desired size, in bytes */
+ int min_size; /* minimum acceptable size (0 means req_size) */
+ long size; /* size of data portion of partition */
+ long index; /* offset of data portion of partition */
+};
+
+static struct nvram_os_partition rtas_log_partition = {
+ .name = "ibm,rtas-log",
+ .req_size = 2079,
+ .min_size = 1055,
+ .index = -1
+};
+
+static const char *pseries_nvram_os_partitions[] = {
+ "ibm,rtas-log",
+ NULL
+};
static ssize_t pSeries_nvram_read(char *buf, size_t count, loff_t *index)
{
@@ -134,7 +147,7 @@ static ssize_t pSeries_nvram_get_size(void)
}
-/* nvram_write_error_log
+/* nvram_write_os_partition, nvram_write_error_log
*
* We need to buffer the error logs into nvram to ensure that we have
* the failure information to decode. If we have a severe error there
@@ -156,48 +169,55 @@ static ssize_t pSeries_nvram_get_size(void)
* The 'data' section would look like (in bytes):
* +--------------+------------+-----------------------------------+
* | event_logged | sequence # | error log |
- * |0 3|4 7|8 nvram_error_log_size-1|
+ * |0 3|4 7|8 error_log_size-1|
* +--------------+------------+-----------------------------------+
*
* event_logged: 0 if event has not been logged to syslog, 1 if it has
* sequence #: The unique sequence # for each event. (until it wraps)
* error log: The error log from event_scan
*/
-int nvram_write_error_log(char * buff, int length,
- unsigned int err_type, unsigned int error_log_cnt)
+int nvram_write_os_partition(struct nvram_os_partition *part, char * buff,
+ int length, unsigned int err_type, unsigned int error_log_cnt)
{
int rc;
loff_t tmp_index;
struct err_log_info info;
- if (nvram_error_log_index == -1) {
+ if (part->index == -1) {
return -ESPIPE;
}
- if (length > nvram_error_log_size) {
- length = nvram_error_log_size;
+ if (length > part->size) {
+ length = part->size;
}
info.error_type = err_type;
info.seq_num = error_log_cnt;
- tmp_index = nvram_error_log_index;
+ tmp_index = part->index;
rc = ppc_md.nvram_write((char *)&info, sizeof(struct err_log_info), &tmp_index);
if (rc <= 0) {
- printk(KERN_ERR "nvram_write_error_log: Failed nvram_write (%d)\n", rc);
+ pr_err("%s: Failed nvram_write (%d)\n", __FUNCTION__, rc);
return rc;
}
rc = ppc_md.nvram_write(buff, length, &tmp_index);
if (rc <= 0) {
- printk(KERN_ERR "nvram_write_error_log: Failed nvram_write (%d)\n", rc);
+ pr_err("%s: Failed nvram_write (%d)\n", __FUNCTION__, rc);
return rc;
}
return 0;
}
+int nvram_write_error_log(char * buff, int length,
+ unsigned int err_type, unsigned int error_log_cnt)
+{
+ return nvram_write_os_partition(&rtas_log_partition, buff, length,
+ err_type, error_log_cnt);
+}
+
/* nvram_read_error_log
*
* Reads nvram for error log for at most 'length'
@@ -209,13 +229,13 @@ int nvram_read_error_log(char * buff, int length,
loff_t tmp_index;
struct err_log_info info;
- if (nvram_error_log_index == -1)
+ if (rtas_log_partition.index == -1)
return -1;
- if (length > nvram_error_log_size)
- length = nvram_error_log_size;
+ if (length > rtas_log_partition.size)
+ length = rtas_log_partition.size;
- tmp_index = nvram_error_log_index;
+ tmp_index = rtas_log_partition.index;
rc = ppc_md.nvram_read((char *)&info, sizeof(struct err_log_info), &tmp_index);
if (rc <= 0) {
@@ -244,10 +264,10 @@ int nvram_clear_error_log(void)
int clear_word = ERR_FLAG_ALREADY_LOGGED;
int rc;
- if (nvram_error_log_index == -1)
+ if (rtas_log_partition.index == -1)
return -1;
- tmp_index = nvram_error_log_index;
+ tmp_index = rtas_log_partition.index;
rc = ppc_md.nvram_write((char *)&clear_word, sizeof(int), &tmp_index);
if (rc <= 0) {
@@ -258,23 +278,25 @@ int nvram_clear_error_log(void)
return 0;
}
-/* pseries_nvram_init_log_partition
+/* pseries_nvram_init_os_partition
*
- * This will setup the partition we need for buffering the
- * error logs and cleanup partitions if needed.
+ * This sets up a partition with an "OS" signature.
*
* The general strategy is the following:
- * 1.) If there is log partition large enough then use it.
- * 2.) If there is none large enough, search
- * for a free partition that is large enough.
- * 3.) If there is not a free partition large enough remove
- * _all_ OS partitions and consolidate the space.
- * 4.) Will first try getting a chunk that will satisfy the maximum
- * error log size (NVRAM_MAX_REQ).
- * 5.) If the max chunk cannot be allocated then try finding a chunk
- * that will satisfy the minum needed (NVRAM_MIN_REQ).
+ * 1.) If a partition with the indicated name already exists...
+ * - If it's large enough, use it.
+ * - Otherwise, recycle it and keep going.
+ * 2.) Search for a free partition that is large enough.
+ * 3.) If there's not a free partition large enough, recycle any obsolete
+ * OS partitions and try again.
+ * 4.) Will first try getting a chunk that will satisfy the requested size.
+ * 5.) If a chunk of the requested size cannot be allocated, then try finding
+ * a chunk that will satisfy the minum needed.
+ *
+ * Returns 0 on success, else -1.
*/
-static int __init pseries_nvram_init_log_partition(void)
+static int __init pseries_nvram_init_os_partition(struct nvram_os_partition
+ *part)
{
loff_t p;
int size;
@@ -282,47 +304,50 @@ static int __init pseries_nvram_init_log_partition(void)
/* Scan nvram for partitions */
nvram_scan_partitions();
- /* Lookg for ours */
- p = nvram_find_partition(NVRAM_LOG_PART_NAME, NVRAM_SIG_OS, &size);
+ /* Look for ours */
+ p = nvram_find_partition(part->name, NVRAM_SIG_OS, &size);
/* Found one but too small, remove it */
- if (p && size < NVRAM_MIN_REQ) {
- pr_info("nvram: Found too small "NVRAM_LOG_PART_NAME" partition"
- ",removing it...");
- nvram_remove_partition(NVRAM_LOG_PART_NAME, NVRAM_SIG_OS);
+ if (p && size < part->min_size) {
+ pr_info("nvram: Found too small %s partition,"
+ " removing it...\n", part->name);
+ nvram_remove_partition(part->name, NVRAM_SIG_OS, NULL);
p = 0;
}
/* Create one if we didn't find */
if (!p) {
- p = nvram_create_partition(NVRAM_LOG_PART_NAME, NVRAM_SIG_OS,
- NVRAM_MAX_REQ, NVRAM_MIN_REQ);
- /* No room for it, try to get rid of any OS partition
- * and try again
- */
+ p = nvram_create_partition(part->name, NVRAM_SIG_OS,
+ part->req_size, part->min_size);
if (p == -ENOSPC) {
- pr_info("nvram: No room to create "NVRAM_LOG_PART_NAME
- " partition, deleting all OS partitions...");
- nvram_remove_partition(NULL, NVRAM_SIG_OS);
- p = nvram_create_partition(NVRAM_LOG_PART_NAME,
- NVRAM_SIG_OS, NVRAM_MAX_REQ,
- NVRAM_MIN_REQ);
+ pr_info("nvram: No room to create %s partition, "
+ "deleting any obsolete OS partitions...\n",
+ part->name);
+ nvram_remove_partition(NULL, NVRAM_SIG_OS,
+ pseries_nvram_os_partitions);
+ p = nvram_create_partition(part->name, NVRAM_SIG_OS,
+ part->req_size, part->min_size);
}
}
if (p <= 0) {
- pr_err("nvram: Failed to find or create "NVRAM_LOG_PART_NAME
- " partition, err %d\n", (int)p);
- return 0;
+ pr_err("nvram: Failed to find or create %s"
+ " partition, err %d\n", part->name, (int)p);
+ return -1;
}
- nvram_error_log_index = p;
- nvram_error_log_size = nvram_get_partition_size(p) -
- sizeof(struct err_log_info);
+ part->index = p;
+ part->size = nvram_get_partition_size(p) - sizeof(struct err_log_info);
return 0;
}
-machine_arch_initcall(pseries, pseries_nvram_init_log_partition);
+
+static int __init pseries_nvram_init_log_partitions(void)
+{
+ (void) pseries_nvram_init_os_partition(&rtas_log_partition);
+ return 0;
+}
+machine_arch_initcall(pseries, pseries_nvram_init_log_partitions);
int __init pSeries_nvram_init(void)
{
^ permalink raw reply related
* Re: [PATCH 1/2] misc: add CARMA DATA-FPGA Access Driver
From: Dmitry Torokhov @ 2011-02-09 18:27 UTC (permalink / raw)
To: Ira W. Snyder; +Cc: linuxppc-dev, linux-kernel
In-Reply-To: <20110209173532.GB21766@ovro.caltech.edu>
On Wed, Feb 09, 2011 at 09:35:32AM -0800, Ira W. Snyder wrote:
> On Wed, Feb 09, 2011 at 12:33:25AM -0800, Dmitry Torokhov wrote:
> > > +
> > > + /* Warn if we are running in a degraded state, but do not fail */
> > > + if (priv->num_buffers < MAX_DATA_BUFS) {
> > > + dev_warn(priv->dev, "Unable to allocate one second worth of "
> > > + "buffers, using %d buffers instead\n", i);
> >
> > The latest style is not break strings even if they exceed 80 column
> > limit to make sure they are easily greppable.
> >
>
> I usually just follow checkpatch warnings. I'll combine the strings onto
> one line.
Does it still warn? I think they tried to fix it so it recognizes long
messages.
> >
> > OTOH, we can't have more than 1 in-flight buffer, can we? Should
> > we even have a list?
> >
>
> This looks like a good change. I didn't know about list_move_tail()
> before you mentioned it. Good catch.
>
> You are correct that it is not possible to have more than one in flight
> buffer at the same time. It was previously possible in an earlier
> version of the driver. That was before I was forced to come up with the
> ugly IRQ disabling scheme due to the hardware design.
>
> What would you replace the inflight list with? A "struct data_buf
> *inflight;" in the priv structure?
>
Yes. Then one would not worry of it is safe to always mark first element
of in-flight list as "complete"
> >
> > Have you considered enabling the device when someone opens the device
> > node and closing it when last user drops off?
> >
> >
>
> Yes. Unfortunately it is not possible. Blame my userspace software
> engineers, who refused this model of operation.
>
> I must meet the requirement that the device must remain open while the
> DATA-FPGAs are reconfigured. This means that the FPGAs are completely
> powered down, and a new FPGA bitfile is loaded into them.
>
> After a reconfiguration, the data buffer size may change.
>
> The userspace coders were willing to use a sysfs file to enable/disable
> reading from the device, but they were not willing to open/close the
> device each time. It was the best compromise they would accept.
>
> I'm not happy about it either.
I see.
>
> > > +
> > > +/*
> > > + * FPGA Realtime Data Character Device
> > > + */
> > > +
> > > +static int data_open(struct inode *inode, struct file *filp)
> > > +{
> > > + /*
> > > + * The miscdevice layer puts our struct miscdevice into the
> > > + * filp->private_data field. We use this to find our private
> > > + * data and then overwrite it with our own private structure.
> > > + */
> > > + struct fpga_device *priv = container_of(filp->private_data,
> > > + struct fpga_device, miscdev);
> > > + struct fpga_reader *reader;
> > > + int ret;
> > > +
> > > + /* allocate private data */
> > > + reader = kzalloc(sizeof(*reader), GFP_KERNEL);
> > > + if (!reader)
> > > + return -ENOMEM;
> > > +
> > > + reader->priv = priv;
> > > + reader->buf = NULL;
> > > +
> > > + filp->private_data = reader;
> > > + ret = nonseekable_open(inode, filp);
> > > + if (ret) {
> > > + dev_err(priv->dev, "nonseekable-open failed\n");
> > > + kfree(reader);
> > > + return ret;
> > > + }
> > > +
> >
> > I wonder if the device should allow only exclusive open... Unless you
> > distribute copies of data between all readers.
> >
>
> I wish to allow multiple different processes to mmap() the device. This
> requires open(), followed by mmap(). I only care about a single realtime
> data reader (which calls read()). Splitting the two use cases into two
> drivers seemed like a big waste of time and effort. I have no idea how
> to accomplish a single read()er while allowing multiple mmap()ers.
I see.
>
> > > + return 0;
> > > +}
> > > +
> > > +static int data_release(struct inode *inode, struct file *filp)
> > > +{
> > > + struct fpga_reader *reader = filp->private_data;
> > > +
> > > + /* free the per-reader structure */
> > > + data_free_buffer(reader->buf);
> > > + kfree(reader);
> > > + filp->private_data = NULL;
> > > + return 0;
> > > +}
> > > +
> > > +static ssize_t data_read(struct file *filp, char __user *ubuf, size_t count,
> > > + loff_t *f_pos)
> > > +{
> > > + struct fpga_reader *reader = filp->private_data;
> > > + struct fpga_device *priv = reader->priv;
> > > + struct list_head *used = &priv->used;
> > > + struct data_buf *dbuf;
> > > + size_t avail;
> > > + void *data;
> > > + int ret;
> > > +
> > > + /* check if we already have a partial buffer */
> > > + if (reader->buf) {
> > > + dbuf = reader->buf;
> > > + goto have_buffer;
> > > + }
> > > +
> > > + spin_lock_irq(&priv->lock);
> > > +
> > > + /* Block until there is at least one buffer on the used list */
> > > + while (list_empty(used)) {
> >
> > I believe we should stop and return error when device is disabled so the
> > condition should be:
> >
> > while (!reader->buf && list_empty() && priv->enabled)
> >
> >
>
> The requirement is that the device stay open during reconfiguration.
> This provides for that. Readers just block for as long as the device is
> not producing data.
OK, you still need to make sure you do not touch free/used buffer while
device is disabled. Also, you need to kick readers if you unbind the
driver, so maybe a new flag priv->exists should be introduced and
checked.
>
> > > + spin_unlock_irq(&priv->lock);
> > > +
> > > + if (filp->f_flags & O_NONBLOCK)
> > > + return -EAGAIN;
> > > +
> > > + ret = wait_event_interruptible(priv->wait, !list_empty(used));
> >
> > ret = wait_event_interruptible(priv->wait,
> > !list_empty(used) || !priv->enabled);
> > > + if (ret)
> > > + return ret;
> > > +
> > > + spin_lock_irq(&priv->lock);
> > > + }
> > > +
> >
> > if (!priv->enabled) {
> > spin_unlock_irq(&priv->lock);
> > return -ENXIO;
> > }
> >
> > if (reader->buf) {
> > dbuf = reader->buf;
> > } else {
> > dbuf = list_first_entry(used, struct data_buf, entry);
> > list_del_init(&dbuf->entry);
> > }
> >
> > > + /* Grab the first buffer off of the used list */
> > > + dbuf = list_first_entry(used, struct data_buf, entry);
> > > + list_del_init(&dbuf->entry);
> > > +
> > > + spin_unlock_irq(&priv->lock);
> > > +
> > > + /* Buffers are always mapped: unmap it */
> > > + data_unmap_buffer(priv->dev, dbuf);
> > > +
> > > + /* save the buffer for later */
> > > + reader->buf = dbuf;
> > > + reader->buf_start = 0;
> > > +
> > > + /* we removed a buffer from the used list: wake any waiters */
> > > + wake_up(&priv->wait);
> >
> > I do not think anyone waits on this...
> >
> > > +
> > > +have_buffer:
> > > + /* Get the number of bytes available */
> > > + avail = dbuf->size - reader->buf_start;
> > > + data = dbuf->vb.vaddr + reader->buf_start;
> > > +
> > > + /* Get the number of bytes we can transfer */
> > > + count = min(count, avail);
> > > +
> > > + /* Copy the data to the userspace buffer */
> > > + if (copy_to_user(ubuf, data, count))
> > > + return -EFAULT;
> > > +
> > > + /* Update the amount of available space */
> > > + avail -= count;
> > > +
> > > + /* Lock against concurrent enable/disable */
> > > + ret = mutex_lock_interruptible(&priv->mutex);
> > > + if (ret)
> > > + return ret;
> >
> > Mutex is not needed here, we shall rely on priv->lock. Map buffer
> > beforehand, take lock, check if the device is enabled and if it still is
> > put the buffer on free list. Release lock and exit if device was
> > enabled; otherwise dispose the buffer.
> >
> > > +
> > > + /* Still some space available: save the buffer for later */
> > > + if (avail != 0) {
> > > + reader->buf_start += count;
> > > + reader->buf = dbuf;
> > > + goto out_unlock;
> > > + }
> > > +
> > > + /*
> > > + * No space is available in this buffer
> > > + *
> > > + * This is a complicated decision:
> > > + * - if the device is not enabled: free the buffer
> > > + * - if the buffer is too small: free the buffer
> > > + */
> > > + if (!priv->enabled || dbuf->size != priv->bufsize) {
> > > + data_free_buffer(dbuf);
> > > + reader->buf = NULL;
> > > + goto out_unlock;
> > > + }
> > > +
> > > + /*
> > > + * The buffer is safe to recycle: remap it and finish
> > > + *
> > > + * If this fails, we pretend that the read never happened, and return
> > > + * -EFAULT to userspace. They'll retry the read again.
> > > + */
> > > + ret = data_map_buffer(priv->dev, dbuf);
> > > + if (ret) {
> > > + dev_err(priv->dev, "unable to remap buffer for DMA\n");
> > > + count = -EFAULT;
> > > + goto out_unlock;
> > > + }
> > > +
> > > + /* Add the buffer back to the free list */
> > > + reader->buf = NULL;
> > > + spin_lock_irq(&priv->lock);
> > > + list_add_tail(&dbuf->entry, &priv->free);
> > > + spin_unlock_irq(&priv->lock);
> > > +
> > > +out_unlock:
> > > + mutex_unlock(&priv->mutex);
> > > + return count;
> > > +}
> > > +
> > > +static unsigned int data_poll(struct file *filp, struct poll_table_struct *tbl)
> > > +{
> > > + struct fpga_reader *reader = filp->private_data;
> > > + struct fpga_device *priv = reader->priv;
> > > + unsigned int mask = 0;
> > > +
> > > + poll_wait(filp, &priv->wait, tbl);
> > > +
> > > + spin_lock_irq(&priv->lock);
> > > +
> >
> > Like I mentioned, the spinlock is not useful here - all threads will get
> > woken up and will try to read since you are not resetting the wakeup
> > condition.
> >
>
> Whoops, I forgot this from your previous review. Sorry.
>
> > > + if (!list_empty(&priv->used))
> > > + mask |= POLLIN | POLLRDNORM;
> >
> > I think you should also add:
> >
> > if (!priv->)
> > mask |= POLLHUP | POLLERR;
> >
> > to tell the readers that they should close their file descriptors.
> >
>
> Did you mean "!priv->enabled"? If so, see the comments above about
> keeping the device open across reconfiguration.
The new !priv->exists then.
Thanks.
--
Dmitry
^ permalink raw reply
* Re: [PATCH 1/2] misc: add CARMA DATA-FPGA Access Driver
From: Ira W. Snyder @ 2011-02-09 17:35 UTC (permalink / raw)
To: Dmitry Torokhov; +Cc: linuxppc-dev, linux-kernel
In-Reply-To: <20110209083325.GA7256@core.coreip.homeip.net>
On Wed, Feb 09, 2011 at 12:33:25AM -0800, Dmitry Torokhov wrote:
> Hi Ira,
>
> On Tue, Feb 08, 2011 at 03:37:46PM -0800, Ira W. Snyder wrote:
> > This driver allows userspace to access the data processing FPGAs on the
> > OVRO CARMA board. It has two modes of operation:
> >
>
> Thank you for making the changes, some more comments below.
>
> > +
> > +#define inode_to_dev(inode) container_of(inode->i_cdev, struct fpga_device, cdev)
> > +
>
> Does not seem to be used.
>
Leftovers from earlier versions, will remove.
> > +/*
> > + * Data Buffer Allocation Helpers
> > + */
> > +
> > +static int data_map_buffer(struct device *dev, struct data_buf *buf)
> > +{
> > + return videobuf_dma_map(dev, &buf->vb);
> > +}
> > +
> > +static void data_unmap_buffer(struct device *dev, struct data_buf *buf)
> > +{
> > + videobuf_dma_unmap(dev, &buf->vb);
> > +}
>
> Why can't we all videobuf_dma_map() and videobuf_dma_unmap() directly?
> What these helpers supposed to solve?
>
The helpers were useful in the past. Now they are not. Will change.
> > +static int data_alloc_buffers(struct fpga_device *priv)
> > +{
> > + struct data_buf *buf;
> > + int i, ret;
> > +
> > + for (i = 0; i < MAX_DATA_BUFS; i++) {
> > +
> > + /* allocate a buffer */
> > + buf = data_alloc_buffer(priv->bufsize);
> > + if (!buf)
> > + continue;
>
> break?
>
> > +
> > + /* map it for DMA */
> > + ret = data_map_buffer(priv->dev, buf);
> > + if (ret) {
> > + data_free_buffer(buf);
> > + continue;
>
> and here as well?
>
Yep, break is fine also.
> > + }
> > +
> > + /* add it to the list of free buffers */
> > + list_add_tail(&buf->entry, &priv->free);
> > + priv->num_buffers++;
> > + }
> > +
> > + /* Make sure we allocated the minimum required number of buffers */
> > + if (priv->num_buffers < MIN_DATA_BUFS) {
> > + dev_err(priv->dev, "Unable to allocate enough data buffers\n");
> > + data_free_buffers(priv);
> > + return -ENOMEM;
> > + }
> > +
> > + /* Warn if we are running in a degraded state, but do not fail */
> > + if (priv->num_buffers < MAX_DATA_BUFS) {
> > + dev_warn(priv->dev, "Unable to allocate one second worth of "
> > + "buffers, using %d buffers instead\n", i);
>
> The latest style is not break strings even if they exceed 80 column
> limit to make sure they are easily greppable.
>
I usually just follow checkpatch warnings. I'll combine the strings onto
one line.
> > +static void data_dma_cb(void *data)
> > +{
> > + struct fpga_device *priv = data;
> > + struct data_buf *buf;
> > + unsigned long flags;
> > +
> > + spin_lock_irqsave(&priv->lock, flags);
> > +
> > + /* clear the FPGA status and re-enable interrupts */
> > + data_enable_interrupts(priv);
> > +
> > + /* If the inflight list is empty, we've got a bug */
> > + BUG_ON(list_empty(&priv->inflight));
> > +
> > + /* Grab the first buffer from the inflight list */
> > + buf = list_first_entry(&priv->inflight, struct data_buf, entry);
> > + list_del_init(&buf->entry);
> > +
> > + /* Add it to the used list */
> > + list_add_tail(&buf->entry, &priv->used);
>
> or
> list_move_tail(&buf->entry, &priv->used);
>
Using list_move_tail() didn't occur to me. I'll change it.
> > +
> > +static irqreturn_t data_irq(int irq, void *dev_id)
> > +{
> > + struct fpga_device *priv = dev_id;
> > + struct data_buf *buf;
> > + u32 status;
> > + int i;
> > +
> > + /* detect spurious interrupts via FPGA status */
> > + for (i = 0; i < 4; i++) {
> > + status = fpga_read_reg(priv, i, MMAP_REG_STATUS);
> > + if (!(status & (CORL_DONE | CORL_ERR))) {
> > + dev_err(priv->dev, "spurious irq detected (FPGA)\n");
> > + return IRQ_NONE;
> > + }
> > + }
> > +
> > + /* detect spurious interrupts via raw IRQ pin readback */
> > + status = ioread32be(priv->regs + SYS_IRQ_INPUT_DATA);
> > + if (status & IRQ_CORL_DONE) {
> > + dev_err(priv->dev, "spurious irq detected (IRQ)\n");
> > + return IRQ_NONE;
> > + }
> > +
> > + spin_lock(&priv->lock);
> > +
> > + /* hide the interrupt by switching the IRQ driver to GPIO */
> > + data_disable_interrupts(priv);
> > +
> > + /* Check that we actually have a free buffer */
> > + if (list_empty(&priv->free)) {
> > + priv->num_dropped++;
> > + data_enable_interrupts(priv);
> > + goto out_unlock;
> > + }
> > +
> > + buf = list_first_entry(&priv->free, struct data_buf, entry);
> > + list_del_init(&buf->entry);
> > +
> > + /* Check the buffer size */
> > + BUG_ON(buf->size != priv->bufsize);
> > +
> > + /* Submit a DMA transfer to get the correlation data */
> > + if (data_submit_dma(priv, buf)) {
> > + dev_err(priv->dev, "Unable to setup DMA transfer\n");
> > + list_add_tail(&buf->entry, &priv->free);
> > + data_enable_interrupts(priv);
> > + goto out_unlock;
> > + }
> > +
> > + /* DMA setup succeeded, GO!!! */
> > + list_add_tail(&buf->entry, &priv->inflight);
> > + dma_async_memcpy_issue_pending(priv->chan);
> > +
> > +out_unlock:
> > + spin_unlock(&priv->lock);
> > + return IRQ_HANDLED;
> > +}
>
> Hmm, I wonder if we could simplify this a bit by using list_move()...
>
> bool submitted = false;
> ...
>
> if (list_empty(&priv->free)) {
> priv->num_dropped++;
> goto out;
> }
>
> buf = list_first_entry(&priv->free, struct data_buf, entry);
> BUG_ON(buf->size != priv->bufsize);
>
> /* Submit a DMA transfer to get the correlation data */
> if (data_submit_dma(priv, buf)) {
> dev_err(priv->dev, "Unable to setup DMA transfer\n");
> goto out;
> }
>
> /* DMA setup succeeded, GO!!! */
> list_move_tail(&buf->entry, &priv->inflight);
> dma_async_memcpy_issue_pending(priv->chan);
> submitted = true;
>
> out:
> if (!submitted)
> data_enable_interrupts(priv);
> spin_unlock(&priv->lock);
> return IRQ_HANDLED;
> }
>
> OTOH, we can't have more than 1 in-flight buffer, can we? Should
> we even have a list?
>
This looks like a good change. I didn't know about list_move_tail()
before you mentioned it. Good catch.
You are correct that it is not possible to have more than one in flight
buffer at the same time. It was previously possible in an earlier
version of the driver. That was before I was forced to come up with the
ugly IRQ disabling scheme due to the hardware design.
What would you replace the inflight list with? A "struct data_buf
*inflight;" in the priv structure?
> > +
> > +/*
> > + * Realtime Device Enable Helpers
> > + */
> > +
> > +/**
> > + * data_device_enable() - enable the device for buffered dumping
> > + * @priv: the driver's private data structure
> > + *
> > + * Enable the device for buffered dumping. Allocates buffers and hooks up
> > + * the interrupt handler. When this finishes, data will come pouring in.
> > + *
> > + * LOCKING: must hold dev->mutex
> > + * CONTEXT: user context only
> > + *
> > + * Returns 0 on success, -ERRNO otherwise
> > + */
> > +static int data_device_enable(struct fpga_device *priv)
> > +{
> > + u32 val;
> > + int ret;
> > +
> > + /* multiple enables are safe: they do nothing */
> > + if (priv->enabled)
> > + return 0;
> > +
> > + /* check that the FPGAs are programmed */
> > + val = ioread32be(priv->regs + SYS_FPGA_CONFIG_STATUS);
> > + if (!(val & (1 << 18))) {
> > + dev_err(priv->dev, "DATA-FPGAs are not enabled\n");
> > + return -ENODATA;
> > + }
> > +
> > + /* read the FPGAs to calculate the buffer size */
> > + ret = data_calculate_bufsize(priv);
> > + if (ret) {
> > + dev_err(priv->dev, "unable to calculate buffer size\n");
> > + goto out_error;
> > + }
> > +
> > + /* allocate the correlation data buffers */
> > + ret = data_alloc_buffers(priv);
> > + if (ret) {
> > + dev_err(priv->dev, "unable to allocate buffers\n");
> > + goto out_error;
> > + }
> > +
> > + /* setup the source scatterlist for dumping correlation data */
> > + ret = data_setup_corl_table(priv);
> > + if (ret) {
> > + dev_err(priv->dev, "unable to setup correlation DMA table\n");
> > + goto out_error;
> > + }
> > +
> > + /* switch to the external FPGA IRQ line */
> > + data_enable_interrupts(priv);
>
> Not after setting interrupt handler? Can't that lead to "unhandled
> interrupt" scenarios?
>
Yes, good catch.
> > +
> > + /* hookup the irq handler */
> > + ret = request_irq(priv->irq, data_irq, IRQF_SHARED, drv_name, priv);
> > + if (ret) {
> > + dev_err(priv->dev, "unable to request IRQ handler\n");
> > + goto out_error;
> > + }
> > +
> > + /* success, we're enabled */
> > + priv->enabled = true;
> > + return 0;
> > +
> > +out_error:
> > + sg_free_table(&priv->corl_table);
> > + priv->corl_nents = 0;
> > +
> > + data_free_buffers(priv);
> > + return ret;
> > +}
> > +
> > +/**
> > + * data_device_disable() - disable the device for buffered dumping
> > + * @priv: the driver's private data structure
> > + *
> > + * Disable the device for buffered dumping. Stops new DMA transactions from
> > + * being generated, waits for all outstanding DMA to complete, and then frees
> > + * all buffers.
> > + *
> > + * LOCKING: must hold dev->mutex
> > + * CONTEXT: user only
> > + *
> > + * Returns 0 on success, -ERRNO otherwise
> > + */
> > +static int data_device_disable(struct fpga_device *priv)
> > +{
> > + struct list_head *list;
> > + int ret;
> > +
> > + /* allow multiple disable */
> > + if (!priv->enabled)
> > + return 0;
> > +
> > + /* switch to the internal GPIO IRQ line */
> > + data_disable_interrupts(priv);
> > +
> > + /* unhook the irq handler */
> > + free_irq(priv->irq, priv);
> > +
> > + /*
> > + * wait for all outstanding DMA to complete
> > + *
> > + * Device interrupts are disabled, so no more buffers can be added to
> > + * the inflight list. Therefore we do not need any locking.
> > + */
> > + list = &priv->inflight;
> > + while (!list_empty(list)) {
> > + ret = wait_event_interruptible(priv->wait, list_empty(list));
> > + if (ret)
> > + return ret;
> > + }
>
> Simply
>
> ret = wait_event_interruptible(priv->wait, list_empty(&prov->inflight));
> if (ret)
> return ret;
>
> is enough, no need to loop.
>
> > +
> > + /* free the correlation table */
> > + sg_free_table(&priv->corl_table);
> > + priv->corl_nents = 0;
> > +
> > + /* free all of the buffers */
> > + data_free_buffers(priv);
> > + priv->enabled = false;
>
> I think this should be:
>
> /*
> * We are taking lock not to protect priv->enabled but to make
> * sure there are no readers in process of altering free/used
> * lists while we are setting the flag.
> */
> spin_lock_irq(&priv->lock);
> priv->enabled = false;
> spin_unlock_irq(&priv->lock);
>
> /*
> * Wake up readers so they error out and hopefully close their
> * file descriptors.
> */
> wake_up(&priv->wait);
>
> /* We now know that readers will not touch free/used lists */
> data_free_buffers();
>
> Also please see the comments in data_read().
>
>
Yes, looks good.
> > + return 0;
> > +}
> > +
> > +/*
> > + * SYSFS Attributes
> > + */
> > +
> > +/*
> > + * Count the number of entries in the given list
> > + */
> > +static unsigned int list_num_entries(struct list_head *list)
> > +{
> > + struct list_head *entry;
> > + unsigned int ret = 0;
> > +
> > + list_for_each(entry, list)
> > + ret++;
> > +
> > + return ret;
> > +}
> > +
> > +static ssize_t data_num_buffers_show(struct device *dev,
> > + struct device_attribute *attr, char *buf)
> > +{
> > + struct fpga_device *priv = dev_get_drvdata(dev);
> > + return snprintf(buf, PAGE_SIZE, "%u\n", priv->num_buffers);
> > +}
> > +
> > +static ssize_t data_bufsize_show(struct device *dev,
> > + struct device_attribute *attr, char *buf)
> > +{
> > + struct fpga_device *priv = dev_get_drvdata(dev);
> > + return snprintf(buf, PAGE_SIZE, "%zu\n", priv->bufsize);
> > +}
> > +
> > +static ssize_t data_inflight_show(struct device *dev,
> > + struct device_attribute *attr, char *buf)
> > +{
> > + struct fpga_device *priv = dev_get_drvdata(dev);
> > + unsigned int num = list_num_entries(&priv->inflight);
> > + return snprintf(buf, PAGE_SIZE, "%u\n", num);
> > +}
> > +
> > +static ssize_t data_free_show(struct device *dev,
> > + struct device_attribute *attr, char *buf)
> > +{
> > + struct fpga_device *priv = dev_get_drvdata(dev);
> > + unsigned int num = list_num_entries(&priv->free);
> > + return snprintf(buf, PAGE_SIZE, "%u\n", num);
> > +}
> > +
> > +static ssize_t data_used_show(struct device *dev,
> > + struct device_attribute *attr, char *buf)
> > +{
> > + struct fpga_device *priv = dev_get_drvdata(dev);
> > + unsigned int num = list_num_entries(&priv->used);
> > + return snprintf(buf, PAGE_SIZE, "%u\n", num);
> > +}
> > +
> > +static ssize_t data_num_dropped_show(struct device *dev,
> > + struct device_attribute *attr, char *buf)
> > +{
> > + struct fpga_device *priv = dev_get_drvdata(dev);
> > + return snprintf(buf, PAGE_SIZE, "%u\n", priv->num_dropped);
> > +}
> > +
> > +static ssize_t data_en_show(struct device *dev, struct device_attribute *attr,
> > + char *buf)
> > +{
> > + struct fpga_device *priv = dev_get_drvdata(dev);
> > + ssize_t count;
> > +
> > + count = mutex_lock_interruptible(&priv->mutex);
> > + if (count)
> > + return count;
> > +
> > + count = snprintf(buf, PAGE_SIZE, "%u\n", priv->enabled);
> > + mutex_unlock(&priv->mutex);
>
> No locking needed.
>
> > + return count;
> > +}
> > +
> > +static ssize_t data_en_set(struct device *dev, struct device_attribute *attr,
> > + const char *buf, size_t count)
> > +{
> > + struct fpga_device *priv = dev_get_drvdata(dev);
> > + unsigned long enable;
> > + int ret;
> > +
> > + ret = strict_strtoul(buf, 0, &enable);
> > + if (ret) {
> > + dev_err(priv->dev, "unable to parse enable input\n");
> > + return -EINVAL;
> > + }
> > +
> > + ret = mutex_lock_interruptible(&priv->mutex);
> > + if (ret)
> > + return ret;
> > +
> > + if (enable)
> > + ret = data_device_enable(priv);
> > + else
> > + ret = data_device_disable(priv);
> > +
>
> Have you considered enabling the device when someone opens the device
> node and closing it when last user drops off?
>
>
Yes. Unfortunately it is not possible. Blame my userspace software
engineers, who refused this model of operation.
I must meet the requirement that the device must remain open while the
DATA-FPGAs are reconfigured. This means that the FPGAs are completely
powered down, and a new FPGA bitfile is loaded into them.
After a reconfiguration, the data buffer size may change.
The userspace coders were willing to use a sysfs file to enable/disable
reading from the device, but they were not willing to open/close the
device each time. It was the best compromise they would accept.
I'm not happy about it either.
> > + if (ret) {
> > + dev_err(priv->dev, "device %s failed\n",
> > + enable ? "enable" : "disable");
> > + count = ret;
> > + goto out_unlock;
> > + }
> > +
> > +out_unlock:
> > + mutex_unlock(&priv->mutex);
> > + return count;
> > +}
> > +
> > +static DEVICE_ATTR(num_buffers, S_IRUGO, data_num_buffers_show, NULL);
> > +static DEVICE_ATTR(buffer_size, S_IRUGO, data_bufsize_show, NULL);
> > +static DEVICE_ATTR(num_inflight, S_IRUGO, data_inflight_show, NULL);
> > +static DEVICE_ATTR(num_free, S_IRUGO, data_free_show, NULL);
> > +static DEVICE_ATTR(num_used, S_IRUGO, data_used_show, NULL);
> > +static DEVICE_ATTR(num_dropped, S_IRUGO, data_num_dropped_show, NULL);
> > +static DEVICE_ATTR(enable, S_IWUSR | S_IRUGO, data_en_show, data_en_set);
> > +
> > +static struct attribute *data_sysfs_attrs[] = {
> > + &dev_attr_num_buffers.attr,
> > + &dev_attr_buffer_size.attr,
> > + &dev_attr_num_inflight.attr,
> > + &dev_attr_num_free.attr,
> > + &dev_attr_num_used.attr,
> > + &dev_attr_num_dropped.attr,
> > + &dev_attr_enable.attr,
> > + NULL,
> > +};
> > +
> > +static const struct attribute_group rt_sysfs_attr_group = {
> > + .attrs = data_sysfs_attrs,
> > +};
>
> As I mentioned, consider switching to debugfs - you won't need multiple
> attributes and will be able to get full picture (number of
> free/used/in-flight/etc) in one file and will not be constrained with
> ABI concerns.
>
I will look into it.
> > +
> > +/*
> > + * FPGA Realtime Data Character Device
> > + */
> > +
> > +static int data_open(struct inode *inode, struct file *filp)
> > +{
> > + /*
> > + * The miscdevice layer puts our struct miscdevice into the
> > + * filp->private_data field. We use this to find our private
> > + * data and then overwrite it with our own private structure.
> > + */
> > + struct fpga_device *priv = container_of(filp->private_data,
> > + struct fpga_device, miscdev);
> > + struct fpga_reader *reader;
> > + int ret;
> > +
> > + /* allocate private data */
> > + reader = kzalloc(sizeof(*reader), GFP_KERNEL);
> > + if (!reader)
> > + return -ENOMEM;
> > +
> > + reader->priv = priv;
> > + reader->buf = NULL;
> > +
> > + filp->private_data = reader;
> > + ret = nonseekable_open(inode, filp);
> > + if (ret) {
> > + dev_err(priv->dev, "nonseekable-open failed\n");
> > + kfree(reader);
> > + return ret;
> > + }
> > +
>
> I wonder if the device should allow only exclusive open... Unless you
> distribute copies of data between all readers.
>
I wish to allow multiple different processes to mmap() the device. This
requires open(), followed by mmap(). I only care about a single realtime
data reader (which calls read()). Splitting the two use cases into two
drivers seemed like a big waste of time and effort. I have no idea how
to accomplish a single read()er while allowing multiple mmap()ers.
> > + return 0;
> > +}
> > +
> > +static int data_release(struct inode *inode, struct file *filp)
> > +{
> > + struct fpga_reader *reader = filp->private_data;
> > +
> > + /* free the per-reader structure */
> > + data_free_buffer(reader->buf);
> > + kfree(reader);
> > + filp->private_data = NULL;
> > + return 0;
> > +}
> > +
> > +static ssize_t data_read(struct file *filp, char __user *ubuf, size_t count,
> > + loff_t *f_pos)
> > +{
> > + struct fpga_reader *reader = filp->private_data;
> > + struct fpga_device *priv = reader->priv;
> > + struct list_head *used = &priv->used;
> > + struct data_buf *dbuf;
> > + size_t avail;
> > + void *data;
> > + int ret;
> > +
> > + /* check if we already have a partial buffer */
> > + if (reader->buf) {
> > + dbuf = reader->buf;
> > + goto have_buffer;
> > + }
> > +
> > + spin_lock_irq(&priv->lock);
> > +
> > + /* Block until there is at least one buffer on the used list */
> > + while (list_empty(used)) {
>
> I believe we should stop and return error when device is disabled so the
> condition should be:
>
> while (!reader->buf && list_empty() && priv->enabled)
>
>
The requirement is that the device stay open during reconfiguration.
This provides for that. Readers just block for as long as the device is
not producing data.
> > + spin_unlock_irq(&priv->lock);
> > +
> > + if (filp->f_flags & O_NONBLOCK)
> > + return -EAGAIN;
> > +
> > + ret = wait_event_interruptible(priv->wait, !list_empty(used));
>
> ret = wait_event_interruptible(priv->wait,
> !list_empty(used) || !priv->enabled);
> > + if (ret)
> > + return ret;
> > +
> > + spin_lock_irq(&priv->lock);
> > + }
> > +
>
> if (!priv->enabled) {
> spin_unlock_irq(&priv->lock);
> return -ENXIO;
> }
>
> if (reader->buf) {
> dbuf = reader->buf;
> } else {
> dbuf = list_first_entry(used, struct data_buf, entry);
> list_del_init(&dbuf->entry);
> }
>
> > + /* Grab the first buffer off of the used list */
> > + dbuf = list_first_entry(used, struct data_buf, entry);
> > + list_del_init(&dbuf->entry);
> > +
> > + spin_unlock_irq(&priv->lock);
> > +
> > + /* Buffers are always mapped: unmap it */
> > + data_unmap_buffer(priv->dev, dbuf);
> > +
> > + /* save the buffer for later */
> > + reader->buf = dbuf;
> > + reader->buf_start = 0;
> > +
> > + /* we removed a buffer from the used list: wake any waiters */
> > + wake_up(&priv->wait);
>
> I do not think anyone waits on this...
>
> > +
> > +have_buffer:
> > + /* Get the number of bytes available */
> > + avail = dbuf->size - reader->buf_start;
> > + data = dbuf->vb.vaddr + reader->buf_start;
> > +
> > + /* Get the number of bytes we can transfer */
> > + count = min(count, avail);
> > +
> > + /* Copy the data to the userspace buffer */
> > + if (copy_to_user(ubuf, data, count))
> > + return -EFAULT;
> > +
> > + /* Update the amount of available space */
> > + avail -= count;
> > +
> > + /* Lock against concurrent enable/disable */
> > + ret = mutex_lock_interruptible(&priv->mutex);
> > + if (ret)
> > + return ret;
>
> Mutex is not needed here, we shall rely on priv->lock. Map buffer
> beforehand, take lock, check if the device is enabled and if it still is
> put the buffer on free list. Release lock and exit if device was
> enabled; otherwise dispose the buffer.
>
> > +
> > + /* Still some space available: save the buffer for later */
> > + if (avail != 0) {
> > + reader->buf_start += count;
> > + reader->buf = dbuf;
> > + goto out_unlock;
> > + }
> > +
> > + /*
> > + * No space is available in this buffer
> > + *
> > + * This is a complicated decision:
> > + * - if the device is not enabled: free the buffer
> > + * - if the buffer is too small: free the buffer
> > + */
> > + if (!priv->enabled || dbuf->size != priv->bufsize) {
> > + data_free_buffer(dbuf);
> > + reader->buf = NULL;
> > + goto out_unlock;
> > + }
> > +
> > + /*
> > + * The buffer is safe to recycle: remap it and finish
> > + *
> > + * If this fails, we pretend that the read never happened, and return
> > + * -EFAULT to userspace. They'll retry the read again.
> > + */
> > + ret = data_map_buffer(priv->dev, dbuf);
> > + if (ret) {
> > + dev_err(priv->dev, "unable to remap buffer for DMA\n");
> > + count = -EFAULT;
> > + goto out_unlock;
> > + }
> > +
> > + /* Add the buffer back to the free list */
> > + reader->buf = NULL;
> > + spin_lock_irq(&priv->lock);
> > + list_add_tail(&dbuf->entry, &priv->free);
> > + spin_unlock_irq(&priv->lock);
> > +
> > +out_unlock:
> > + mutex_unlock(&priv->mutex);
> > + return count;
> > +}
> > +
> > +static unsigned int data_poll(struct file *filp, struct poll_table_struct *tbl)
> > +{
> > + struct fpga_reader *reader = filp->private_data;
> > + struct fpga_device *priv = reader->priv;
> > + unsigned int mask = 0;
> > +
> > + poll_wait(filp, &priv->wait, tbl);
> > +
> > + spin_lock_irq(&priv->lock);
> > +
>
> Like I mentioned, the spinlock is not useful here - all threads will get
> woken up and will try to read since you are not resetting the wakeup
> condition.
>
Whoops, I forgot this from your previous review. Sorry.
> > + if (!list_empty(&priv->used))
> > + mask |= POLLIN | POLLRDNORM;
>
> I think you should also add:
>
> if (!priv->)
> mask |= POLLHUP | POLLERR;
>
> to tell the readers that they should close their file descriptors.
>
Did you mean "!priv->enabled"? If so, see the comments above about
keeping the device open across reconfiguration.
Thanks for the review, I'll start making more improvements now. :)
Ira
^ permalink raw reply
* Re: [PATCH 1/2] misc: add CARMA DATA-FPGA Access Driver
From: Ira W. Snyder @ 2011-02-09 17:03 UTC (permalink / raw)
To: David Laight; +Cc: linuxppc-dev, linux-kernel
In-Reply-To: <AE90C24D6B3A694183C094C60CF0A2F6D8AC45@saturn3.aculab.com>
On Wed, Feb 09, 2011 at 04:30:23PM -0000, David Laight wrote:
>
> > This driver allows userspace to access the data processing
> > FPGAs on the OVRO CARMA board. It has two modes of operation:
> >
> > 1) random access
> >
> > This allows users to poke any DATA-FPGA registers by using mmap to map
> > the address region directly into their memory map.
>
> I needed something similar, but used pread() and pwrite()
> to request the transfers.
> While this does require a system call per transfer, it allows
> the driver to use dma (if available) to speed up the request.
> In my case doing single cycle transfers would be too slow.
>
We initially started with a read()/write() interface for individual
register reads and writes, just like you describe. It turned out that
mmap was plenty fast for our use. I made the decision to ditch all of
the extra code needed to setup and execute the DMA for the much simpler
mmap code.
In our case, going all the way through the DMA engine code just to
transfer 4 bytes is overkill. The local bus is already quite fast, and
we can increase the clock speed if needed.
Ira
^ permalink raw reply
* RE: [PATCH 1/2] misc: add CARMA DATA-FPGA Access Driver
From: David Laight @ 2011-02-09 16:30 UTC (permalink / raw)
To: Ira W. Snyder, linuxppc-dev; +Cc: linux-kernel
In-Reply-To: <1297121021-3522-2-git-send-email-iws@ovro.caltech.edu>
=20
> This driver allows userspace to access the data processing=20
> FPGAs on the OVRO CARMA board. It has two modes of operation:
>=20
> 1) random access
>=20
> This allows users to poke any DATA-FPGA registers by using mmap to map
> the address region directly into their memory map.
I needed something similar, but used pread() and pwrite()
to request the transfers.
While this does require a system call per transfer, it allows
the driver to use dma (if available) to speed up the request.
In my case doing single cycle transfers would be too slow.
David
^ permalink raw reply
* Re: [PATCH] Fix masking of interrupts for 52xx GPT IRQ.
From: Henk Stegeman @ 2011-02-09 10:16 UTC (permalink / raw)
To: Benjamin Herrenschmidt; +Cc: linuxppc-dev
In-Reply-To: <1297033514.14982.6.camel@pasglop>
What I'm trying to achieve is to have an interrupt service routine
called for every rising edge on pin E2 of the MPC5200B.
The only CPU register setup (AFAIK) to let the CPU generate this
interrupt is to use the GPT (general purpose timer) function pin E2 is
attached to.
This function is supported by mpc52xx_gpt.c.
Excerpt from mpc52xx_gpt.c:
* To use the IRQ controller function, the following two properties must
* be added to the device tree node for the gpt device:
* interrupt-controller;
* #interrupt-cells =3D < 1 >;
* The IRQ controller binding only uses one cell to specify the interrupt,
* and the IRQ flags are encoded in the cell. A cell is not used to encode
* the IRQ number because the GPT only has a single IRQ source. For flags,
* a value of '1' means rising edge sensitive and '2' means falling edge.
In my dts it's setup like so:
gpt6: timer@660 { // General Purpose Timer GPT6
in GPIO mode for
SMC4000IO sample irq.
compatible =3D "fsl,mpc5200b-gpt","fsl,mpc5200-gpt";
reg =3D <0x660 0x10>;
interrupts =3D <1 15 0>;
interrupt-controller;
#interrupt-cells =3D <1>;
};
When I put a 512Hz signal on pin E2, I was not getting 512 interrupt
calls per second.
I then found that the GPT's "Interrupt Enable" bit is touched via the
mask/unmask functions.
However when 0, the "Interrupt Enable" bit stops generation of
interrupts by the GPT, so while 0, any edge on pin E2 is lost.
If the GPT had it's own mask bit I would probably have changed the
mask/unmask to use this bit instead.
Now the bit that actually would mask/unmask the interrupt is in the
mask register of the MPC5200B's interrupt controller.
This bit (Bit 15 of the interrupts controller's main_mask register) is
controlled by mpc52xx_main_mask / mpc52xx_main_unmask in
mpc52xx_pic.c.
It does feel odd to have it masked/unmask a level down the cascade, if
the GPT provided more interrupt sources they'd be disabled too, but
that's not the case here, only one interrupt is provided by the GPT.
So that's why I did it anyways. It works, and even if it is sane, it's
confusing and I'd sure like to know of a better way to do this.
Because the old code in the unmask/mask function did enable/disable
and I didn't want to just drop that code, I provided it via the
enable/disable function.
What is wrong by implementing & registering both mask/unmask and
enable/disable for the same irq_chip?
If it is wrong it would be nice to let the kernel print a big fat
warning when this is registered.
Cheers,
Henk
On Mon, Feb 7, 2011 at 12:05 AM, Benjamin Herrenschmidt
<benh@kernel.crashing.org> wrote:
> On Sat, 2011-01-15 at 02:28 +0100, Henk Stegeman wrote:
>> When using the GPT as interrupt controller interupts were missing.
>> It turned out the IrqEn bit of the GPT is not a mask bit, but it also
>> disables interrupt generation. This modification masks interrupt one
>> level down the cascade. Note that masking one level down the cascade
>> is only valid here because the GPT as interrupt ontroller only serves
>> one IRQ.
>
> I'm not too sure here... You shouldn't implemen t both mask/unmask and
> enable/disable on the same irq_chip and certainly not cal
> enable_irq/disable_irq from a mask or an unmask callback...
>
> Now, I'm not familiar with the HW here, can you tell me more about what
> exactly is happening, how things are layed out and what you are trying
> to fix ?
>
> Cheers,
> Ben.
>
>> Signed-off-by: Henk Stegeman <henk.stegeman@gmail.com>
>> ---
>> =A0arch/powerpc/platforms/52xx/mpc52xx_gpt.c | =A0 25 ++++++++++++++++++=
++++---
>> =A01 files changed, 22 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/powerpc/platforms/52xx/mpc52xx_gpt.c b/arch/powerpc/pl=
atforms/52xx/mpc52xx_gpt.c
>> index 6f8ebe1..9ae2045 100644
>> --- a/arch/powerpc/platforms/52xx/mpc52xx_gpt.c
>> +++ b/arch/powerpc/platforms/52xx/mpc52xx_gpt.c
>> @@ -91,7 +91,7 @@ struct mpc52xx_gpt_priv {
>> =A0 =A0 =A0 struct irq_host *irqhost;
>> =A0 =A0 =A0 u32 ipb_freq;
>> =A0 =A0 =A0 u8 wdt_mode;
>> -
>> + =A0 =A0 int cascade_virq;
>> =A0#if defined(CONFIG_GPIOLIB)
>> =A0 =A0 =A0 struct of_gpio_chip of_gc;
>> =A0#endif
>> @@ -136,18 +136,35 @@ DEFINE_MUTEX(mpc52xx_gpt_list_mutex);
>> =A0static void mpc52xx_gpt_irq_unmask(unsigned int virq)
>> =A0{
>> =A0 =A0 =A0 struct mpc52xx_gpt_priv *gpt =3D get_irq_chip_data(virq);
>> +
>> + =A0 =A0 enable_irq(gpt->cascade_virq);
>> +
>> +}
>> +
>> +static void mpc52xx_gpt_irq_mask(unsigned int virq)
>> +{
>> + =A0 =A0 struct mpc52xx_gpt_priv *gpt =3D get_irq_chip_data(virq);
>> +
>> + =A0 =A0 disable_irq(gpt->cascade_virq);
>> +}
>> +
>> +static void mpc52xx_gpt_irq_enable(unsigned int virq)
>> +{
>> + =A0 =A0 struct mpc52xx_gpt_priv *gpt =3D get_irq_chip_data(virq);
>> =A0 =A0 =A0 unsigned long flags;
>>
>> + =A0 =A0 dev_dbg(gpt->dev, "%s %d\n", __func__, virq);
>> =A0 =A0 =A0 spin_lock_irqsave(&gpt->lock, flags);
>> =A0 =A0 =A0 setbits32(&gpt->regs->mode, MPC52xx_GPT_MODE_IRQ_EN);
>> =A0 =A0 =A0 spin_unlock_irqrestore(&gpt->lock, flags);
>> =A0}
>>
>> -static void mpc52xx_gpt_irq_mask(unsigned int virq)
>> +static void mpc52xx_gpt_irq_disable(unsigned int virq)
>> =A0{
>> =A0 =A0 =A0 struct mpc52xx_gpt_priv *gpt =3D get_irq_chip_data(virq);
>> =A0 =A0 =A0 unsigned long flags;
>>
>> + =A0 =A0 dev_dbg(gpt->dev, "%s %d\n", __func__, virq);
>> =A0 =A0 =A0 spin_lock_irqsave(&gpt->lock, flags);
>> =A0 =A0 =A0 clrbits32(&gpt->regs->mode, MPC52xx_GPT_MODE_IRQ_EN);
>> =A0 =A0 =A0 spin_unlock_irqrestore(&gpt->lock, flags);
>> @@ -184,6 +201,8 @@ static struct irq_chip mpc52xx_gpt_irq_chip =3D {
>> =A0 =A0 =A0 .name =3D "MPC52xx GPT",
>> =A0 =A0 =A0 .unmask =3D mpc52xx_gpt_irq_unmask,
>> =A0 =A0 =A0 .mask =3D mpc52xx_gpt_irq_mask,
>> + =A0 =A0 .enable =3D mpc52xx_gpt_irq_enable,
>> + =A0 =A0 .disable =3D mpc52xx_gpt_irq_disable,
>> =A0 =A0 =A0 .ack =3D mpc52xx_gpt_irq_ack,
>> =A0 =A0 =A0 .set_type =3D mpc52xx_gpt_irq_set_type,
>> =A0};
>> @@ -268,7 +287,7 @@ mpc52xx_gpt_irq_setup(struct mpc52xx_gpt_priv *gpt, =
struct device_node *node)
>> =A0 =A0 =A0 if ((mode & MPC52xx_GPT_MODE_MS_MASK) =3D=3D 0)
>> =A0 =A0 =A0 =A0 =A0 =A0 =A0 out_be32(&gpt->regs->mode, mode | MPC52xx_GP=
T_MODE_MS_IC);
>> =A0 =A0 =A0 spin_unlock_irqrestore(&gpt->lock, flags);
>> -
>> + =A0 =A0 gpt->cascade_virq =3D cascade_virq;
>> =A0 =A0 =A0 dev_dbg(gpt->dev, "%s() complete. virq=3D%i\n", __func__, ca=
scade_virq);
>> =A0}
>>
>
>
>
^ permalink raw reply
* Re: [PATCH] sched: provide scheduler_ipi() callback in response to smp_send_reschedule()
From: Peter Zijlstra @ 2011-02-09 9:37 UTC (permalink / raw)
To: Benjamin Herrenschmidt
Cc: linux-m32r-ja, linux-mips, linux-ia64, linux-sh, H. Peter Anvin,
Heiko Carstens, Paul Mackerras, Helge Deller, sparclinux,
Linux-Arch, linux-s390, Jesper Nilsson, Jeremy Fitzhardinge,
Russell King, Hirokazu Takata, x86, James E.J. Bottomley,
virtualization, Ingo Molnar, Matt Turner, Fenghua Yu,
Mike Frysinger, user-mode-linux-devel, Konrad Rzeszutek Wilk,
Jeff Dike, Chris Metcalf, xen-devel, Mikael Starvik, linux-m32r,
Ivan Kokshaysky, user-mode-linux-user, uclinux-dist-devel,
Thomas Gleixner, linux-arm-kernel, Richard Henderson, Tony Luck,
linux-parisc, linux-cris-kernel, linux-am33-list, linux-kernel,
Ralf Baechle, Kyle McMartin, Paul Mundt, linux-alpha,
Martin Schwidefsky, linux390, Koichi Yasutake, linuxppc-dev,
David S. Miller
In-Reply-To: <1297232054.14982.346.camel@pasglop>
On Wed, 2011-02-09 at 17:14 +1100, Benjamin Herrenschmidt wrote:
> On Mon, 2011-02-07 at 14:54 +0100, Peter Zijlstra wrote:
> > On Mon, 2011-02-07 at 10:26 +1100, Benjamin Herrenschmidt wrote:
> > > You missed:
> > >
> > > diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
> > > index 9813605..467d122 100644
> > > --- a/arch/powerpc/kernel/smp.c
> > > +++ b/arch/powerpc/kernel/smp.c
> > > @@ -98,6 +98,7 @@ void smp_message_recv(int msg)
> > > break;
> > > case PPC_MSG_RESCHEDULE:
> > > /* we notice need_resched on exit */
> > > + scheduler_ipi();
> > > break;
> > > case PPC_MSG_CALL_FUNC_SINGLE:
> > > generic_smp_call_function_single_interrupt();
> > >
> > > Fold that in and add:
> > >
> > > Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> >
> > Thanks Ben!
>
> BTW. I we lurking at some of our CPU hotplug code and I think I cannot
> totally guarantee that this won't be called on an offline CPU. If that's
> a problem, you may want to add a test for that.
>
> IE. The call function IPIs are normally not going to be sent to an
> offlined CPU, and stop_machine should be a good enough fence here, but
> we do abuse reschedule for a number of things (including in some case
> to wake up a sleeping CPU that was pseudo-offlined :-)
Hmm, I _think_ that should all work out nicely, but we'll see, if when
this stuff hits the tree powerpc machines start falling over we'd better
put that check in ;-)
Meanwhile I'm going to preserve this comment in the changelog of this
patch so we don't forget.
^ permalink raw reply
* Re: [PATCH 1/2] misc: add CARMA DATA-FPGA Access Driver
From: Dmitry Torokhov @ 2011-02-09 8:33 UTC (permalink / raw)
To: Ira W. Snyder; +Cc: linuxppc-dev, linux-kernel
In-Reply-To: <1297208267-27087-2-git-send-email-iws@ovro.caltech.edu>
Hi Ira,
On Tue, Feb 08, 2011 at 03:37:46PM -0800, Ira W. Snyder wrote:
> This driver allows userspace to access the data processing FPGAs on the
> OVRO CARMA board. It has two modes of operation:
>
Thank you for making the changes, some more comments below.
> +
> +#define inode_to_dev(inode) container_of(inode->i_cdev, struct fpga_device, cdev)
> +
Does not seem to be used.
> +/*
> + * Data Buffer Allocation Helpers
> + */
> +
> +static int data_map_buffer(struct device *dev, struct data_buf *buf)
> +{
> + return videobuf_dma_map(dev, &buf->vb);
> +}
> +
> +static void data_unmap_buffer(struct device *dev, struct data_buf *buf)
> +{
> + videobuf_dma_unmap(dev, &buf->vb);
> +}
Why can't we all videobuf_dma_map() and videobuf_dma_unmap() directly?
What these helpers supposed to solve?
> +static int data_alloc_buffers(struct fpga_device *priv)
> +{
> + struct data_buf *buf;
> + int i, ret;
> +
> + for (i = 0; i < MAX_DATA_BUFS; i++) {
> +
> + /* allocate a buffer */
> + buf = data_alloc_buffer(priv->bufsize);
> + if (!buf)
> + continue;
break?
> +
> + /* map it for DMA */
> + ret = data_map_buffer(priv->dev, buf);
> + if (ret) {
> + data_free_buffer(buf);
> + continue;
and here as well?
> + }
> +
> + /* add it to the list of free buffers */
> + list_add_tail(&buf->entry, &priv->free);
> + priv->num_buffers++;
> + }
> +
> + /* Make sure we allocated the minimum required number of buffers */
> + if (priv->num_buffers < MIN_DATA_BUFS) {
> + dev_err(priv->dev, "Unable to allocate enough data buffers\n");
> + data_free_buffers(priv);
> + return -ENOMEM;
> + }
> +
> + /* Warn if we are running in a degraded state, but do not fail */
> + if (priv->num_buffers < MAX_DATA_BUFS) {
> + dev_warn(priv->dev, "Unable to allocate one second worth of "
> + "buffers, using %d buffers instead\n", i);
The latest style is not break strings even if they exceed 80 column
limit to make sure they are easily greppable.
> +static void data_dma_cb(void *data)
> +{
> + struct fpga_device *priv = data;
> + struct data_buf *buf;
> + unsigned long flags;
> +
> + spin_lock_irqsave(&priv->lock, flags);
> +
> + /* clear the FPGA status and re-enable interrupts */
> + data_enable_interrupts(priv);
> +
> + /* If the inflight list is empty, we've got a bug */
> + BUG_ON(list_empty(&priv->inflight));
> +
> + /* Grab the first buffer from the inflight list */
> + buf = list_first_entry(&priv->inflight, struct data_buf, entry);
> + list_del_init(&buf->entry);
> +
> + /* Add it to the used list */
> + list_add_tail(&buf->entry, &priv->used);
or
list_move_tail(&buf->entry, &priv->used);
> +
> +static irqreturn_t data_irq(int irq, void *dev_id)
> +{
> + struct fpga_device *priv = dev_id;
> + struct data_buf *buf;
> + u32 status;
> + int i;
> +
> + /* detect spurious interrupts via FPGA status */
> + for (i = 0; i < 4; i++) {
> + status = fpga_read_reg(priv, i, MMAP_REG_STATUS);
> + if (!(status & (CORL_DONE | CORL_ERR))) {
> + dev_err(priv->dev, "spurious irq detected (FPGA)\n");
> + return IRQ_NONE;
> + }
> + }
> +
> + /* detect spurious interrupts via raw IRQ pin readback */
> + status = ioread32be(priv->regs + SYS_IRQ_INPUT_DATA);
> + if (status & IRQ_CORL_DONE) {
> + dev_err(priv->dev, "spurious irq detected (IRQ)\n");
> + return IRQ_NONE;
> + }
> +
> + spin_lock(&priv->lock);
> +
> + /* hide the interrupt by switching the IRQ driver to GPIO */
> + data_disable_interrupts(priv);
> +
> + /* Check that we actually have a free buffer */
> + if (list_empty(&priv->free)) {
> + priv->num_dropped++;
> + data_enable_interrupts(priv);
> + goto out_unlock;
> + }
> +
> + buf = list_first_entry(&priv->free, struct data_buf, entry);
> + list_del_init(&buf->entry);
> +
> + /* Check the buffer size */
> + BUG_ON(buf->size != priv->bufsize);
> +
> + /* Submit a DMA transfer to get the correlation data */
> + if (data_submit_dma(priv, buf)) {
> + dev_err(priv->dev, "Unable to setup DMA transfer\n");
> + list_add_tail(&buf->entry, &priv->free);
> + data_enable_interrupts(priv);
> + goto out_unlock;
> + }
> +
> + /* DMA setup succeeded, GO!!! */
> + list_add_tail(&buf->entry, &priv->inflight);
> + dma_async_memcpy_issue_pending(priv->chan);
> +
> +out_unlock:
> + spin_unlock(&priv->lock);
> + return IRQ_HANDLED;
> +}
Hmm, I wonder if we could simplify this a bit by using list_move()...
bool submitted = false;
...
if (list_empty(&priv->free)) {
priv->num_dropped++;
goto out;
}
buf = list_first_entry(&priv->free, struct data_buf, entry);
BUG_ON(buf->size != priv->bufsize);
/* Submit a DMA transfer to get the correlation data */
if (data_submit_dma(priv, buf)) {
dev_err(priv->dev, "Unable to setup DMA transfer\n");
goto out;
}
/* DMA setup succeeded, GO!!! */
list_move_tail(&buf->entry, &priv->inflight);
dma_async_memcpy_issue_pending(priv->chan);
submitted = true;
out:
if (!submitted)
data_enable_interrupts(priv);
spin_unlock(&priv->lock);
return IRQ_HANDLED;
}
OTOH, we can't have more than 1 in-flight buffer, can we? Should
we even have a list?
> +
> +/*
> + * Realtime Device Enable Helpers
> + */
> +
> +/**
> + * data_device_enable() - enable the device for buffered dumping
> + * @priv: the driver's private data structure
> + *
> + * Enable the device for buffered dumping. Allocates buffers and hooks up
> + * the interrupt handler. When this finishes, data will come pouring in.
> + *
> + * LOCKING: must hold dev->mutex
> + * CONTEXT: user context only
> + *
> + * Returns 0 on success, -ERRNO otherwise
> + */
> +static int data_device_enable(struct fpga_device *priv)
> +{
> + u32 val;
> + int ret;
> +
> + /* multiple enables are safe: they do nothing */
> + if (priv->enabled)
> + return 0;
> +
> + /* check that the FPGAs are programmed */
> + val = ioread32be(priv->regs + SYS_FPGA_CONFIG_STATUS);
> + if (!(val & (1 << 18))) {
> + dev_err(priv->dev, "DATA-FPGAs are not enabled\n");
> + return -ENODATA;
> + }
> +
> + /* read the FPGAs to calculate the buffer size */
> + ret = data_calculate_bufsize(priv);
> + if (ret) {
> + dev_err(priv->dev, "unable to calculate buffer size\n");
> + goto out_error;
> + }
> +
> + /* allocate the correlation data buffers */
> + ret = data_alloc_buffers(priv);
> + if (ret) {
> + dev_err(priv->dev, "unable to allocate buffers\n");
> + goto out_error;
> + }
> +
> + /* setup the source scatterlist for dumping correlation data */
> + ret = data_setup_corl_table(priv);
> + if (ret) {
> + dev_err(priv->dev, "unable to setup correlation DMA table\n");
> + goto out_error;
> + }
> +
> + /* switch to the external FPGA IRQ line */
> + data_enable_interrupts(priv);
Not after setting interrupt handler? Can't that lead to "unhandled
interrupt" scenarios?
> +
> + /* hookup the irq handler */
> + ret = request_irq(priv->irq, data_irq, IRQF_SHARED, drv_name, priv);
> + if (ret) {
> + dev_err(priv->dev, "unable to request IRQ handler\n");
> + goto out_error;
> + }
> +
> + /* success, we're enabled */
> + priv->enabled = true;
> + return 0;
> +
> +out_error:
> + sg_free_table(&priv->corl_table);
> + priv->corl_nents = 0;
> +
> + data_free_buffers(priv);
> + return ret;
> +}
> +
> +/**
> + * data_device_disable() - disable the device for buffered dumping
> + * @priv: the driver's private data structure
> + *
> + * Disable the device for buffered dumping. Stops new DMA transactions from
> + * being generated, waits for all outstanding DMA to complete, and then frees
> + * all buffers.
> + *
> + * LOCKING: must hold dev->mutex
> + * CONTEXT: user only
> + *
> + * Returns 0 on success, -ERRNO otherwise
> + */
> +static int data_device_disable(struct fpga_device *priv)
> +{
> + struct list_head *list;
> + int ret;
> +
> + /* allow multiple disable */
> + if (!priv->enabled)
> + return 0;
> +
> + /* switch to the internal GPIO IRQ line */
> + data_disable_interrupts(priv);
> +
> + /* unhook the irq handler */
> + free_irq(priv->irq, priv);
> +
> + /*
> + * wait for all outstanding DMA to complete
> + *
> + * Device interrupts are disabled, so no more buffers can be added to
> + * the inflight list. Therefore we do not need any locking.
> + */
> + list = &priv->inflight;
> + while (!list_empty(list)) {
> + ret = wait_event_interruptible(priv->wait, list_empty(list));
> + if (ret)
> + return ret;
> + }
Simply
ret = wait_event_interruptible(priv->wait, list_empty(&prov->inflight));
if (ret)
return ret;
is enough, no need to loop.
> +
> + /* free the correlation table */
> + sg_free_table(&priv->corl_table);
> + priv->corl_nents = 0;
> +
> + /* free all of the buffers */
> + data_free_buffers(priv);
> + priv->enabled = false;
I think this should be:
/*
* We are taking lock not to protect priv->enabled but to make
* sure there are no readers in process of altering free/used
* lists while we are setting the flag.
*/
spin_lock_irq(&priv->lock);
priv->enabled = false;
spin_unlock_irq(&priv->lock);
/*
* Wake up readers so they error out and hopefully close their
* file descriptors.
*/
wake_up(&priv->wait);
/* We now know that readers will not touch free/used lists */
data_free_buffers();
Also please see the comments in data_read().
> + return 0;
> +}
> +
> +/*
> + * SYSFS Attributes
> + */
> +
> +/*
> + * Count the number of entries in the given list
> + */
> +static unsigned int list_num_entries(struct list_head *list)
> +{
> + struct list_head *entry;
> + unsigned int ret = 0;
> +
> + list_for_each(entry, list)
> + ret++;
> +
> + return ret;
> +}
> +
> +static ssize_t data_num_buffers_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct fpga_device *priv = dev_get_drvdata(dev);
> + return snprintf(buf, PAGE_SIZE, "%u\n", priv->num_buffers);
> +}
> +
> +static ssize_t data_bufsize_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct fpga_device *priv = dev_get_drvdata(dev);
> + return snprintf(buf, PAGE_SIZE, "%zu\n", priv->bufsize);
> +}
> +
> +static ssize_t data_inflight_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct fpga_device *priv = dev_get_drvdata(dev);
> + unsigned int num = list_num_entries(&priv->inflight);
> + return snprintf(buf, PAGE_SIZE, "%u\n", num);
> +}
> +
> +static ssize_t data_free_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct fpga_device *priv = dev_get_drvdata(dev);
> + unsigned int num = list_num_entries(&priv->free);
> + return snprintf(buf, PAGE_SIZE, "%u\n", num);
> +}
> +
> +static ssize_t data_used_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct fpga_device *priv = dev_get_drvdata(dev);
> + unsigned int num = list_num_entries(&priv->used);
> + return snprintf(buf, PAGE_SIZE, "%u\n", num);
> +}
> +
> +static ssize_t data_num_dropped_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct fpga_device *priv = dev_get_drvdata(dev);
> + return snprintf(buf, PAGE_SIZE, "%u\n", priv->num_dropped);
> +}
> +
> +static ssize_t data_en_show(struct device *dev, struct device_attribute *attr,
> + char *buf)
> +{
> + struct fpga_device *priv = dev_get_drvdata(dev);
> + ssize_t count;
> +
> + count = mutex_lock_interruptible(&priv->mutex);
> + if (count)
> + return count;
> +
> + count = snprintf(buf, PAGE_SIZE, "%u\n", priv->enabled);
> + mutex_unlock(&priv->mutex);
No locking needed.
> + return count;
> +}
> +
> +static ssize_t data_en_set(struct device *dev, struct device_attribute *attr,
> + const char *buf, size_t count)
> +{
> + struct fpga_device *priv = dev_get_drvdata(dev);
> + unsigned long enable;
> + int ret;
> +
> + ret = strict_strtoul(buf, 0, &enable);
> + if (ret) {
> + dev_err(priv->dev, "unable to parse enable input\n");
> + return -EINVAL;
> + }
> +
> + ret = mutex_lock_interruptible(&priv->mutex);
> + if (ret)
> + return ret;
> +
> + if (enable)
> + ret = data_device_enable(priv);
> + else
> + ret = data_device_disable(priv);
> +
Have you considered enabling the device when someone opens the device
node and closing it when last user drops off?
> + if (ret) {
> + dev_err(priv->dev, "device %s failed\n",
> + enable ? "enable" : "disable");
> + count = ret;
> + goto out_unlock;
> + }
> +
> +out_unlock:
> + mutex_unlock(&priv->mutex);
> + return count;
> +}
> +
> +static DEVICE_ATTR(num_buffers, S_IRUGO, data_num_buffers_show, NULL);
> +static DEVICE_ATTR(buffer_size, S_IRUGO, data_bufsize_show, NULL);
> +static DEVICE_ATTR(num_inflight, S_IRUGO, data_inflight_show, NULL);
> +static DEVICE_ATTR(num_free, S_IRUGO, data_free_show, NULL);
> +static DEVICE_ATTR(num_used, S_IRUGO, data_used_show, NULL);
> +static DEVICE_ATTR(num_dropped, S_IRUGO, data_num_dropped_show, NULL);
> +static DEVICE_ATTR(enable, S_IWUSR | S_IRUGO, data_en_show, data_en_set);
> +
> +static struct attribute *data_sysfs_attrs[] = {
> + &dev_attr_num_buffers.attr,
> + &dev_attr_buffer_size.attr,
> + &dev_attr_num_inflight.attr,
> + &dev_attr_num_free.attr,
> + &dev_attr_num_used.attr,
> + &dev_attr_num_dropped.attr,
> + &dev_attr_enable.attr,
> + NULL,
> +};
> +
> +static const struct attribute_group rt_sysfs_attr_group = {
> + .attrs = data_sysfs_attrs,
> +};
As I mentioned, consider switching to debugfs - you won't need multiple
attributes and will be able to get full picture (number of
free/used/in-flight/etc) in one file and will not be constrained with
ABI concerns.
> +
> +/*
> + * FPGA Realtime Data Character Device
> + */
> +
> +static int data_open(struct inode *inode, struct file *filp)
> +{
> + /*
> + * The miscdevice layer puts our struct miscdevice into the
> + * filp->private_data field. We use this to find our private
> + * data and then overwrite it with our own private structure.
> + */
> + struct fpga_device *priv = container_of(filp->private_data,
> + struct fpga_device, miscdev);
> + struct fpga_reader *reader;
> + int ret;
> +
> + /* allocate private data */
> + reader = kzalloc(sizeof(*reader), GFP_KERNEL);
> + if (!reader)
> + return -ENOMEM;
> +
> + reader->priv = priv;
> + reader->buf = NULL;
> +
> + filp->private_data = reader;
> + ret = nonseekable_open(inode, filp);
> + if (ret) {
> + dev_err(priv->dev, "nonseekable-open failed\n");
> + kfree(reader);
> + return ret;
> + }
> +
I wonder if the device should allow only exclusive open... Unless you
distribute copies of data between all readers.
> + return 0;
> +}
> +
> +static int data_release(struct inode *inode, struct file *filp)
> +{
> + struct fpga_reader *reader = filp->private_data;
> +
> + /* free the per-reader structure */
> + data_free_buffer(reader->buf);
> + kfree(reader);
> + filp->private_data = NULL;
> + return 0;
> +}
> +
> +static ssize_t data_read(struct file *filp, char __user *ubuf, size_t count,
> + loff_t *f_pos)
> +{
> + struct fpga_reader *reader = filp->private_data;
> + struct fpga_device *priv = reader->priv;
> + struct list_head *used = &priv->used;
> + struct data_buf *dbuf;
> + size_t avail;
> + void *data;
> + int ret;
> +
> + /* check if we already have a partial buffer */
> + if (reader->buf) {
> + dbuf = reader->buf;
> + goto have_buffer;
> + }
> +
> + spin_lock_irq(&priv->lock);
> +
> + /* Block until there is at least one buffer on the used list */
> + while (list_empty(used)) {
I believe we should stop and return error when device is disabled so the
condition should be:
while (!reader->buf && list_empty() && priv->enabled)
> + spin_unlock_irq(&priv->lock);
> +
> + if (filp->f_flags & O_NONBLOCK)
> + return -EAGAIN;
> +
> + ret = wait_event_interruptible(priv->wait, !list_empty(used));
ret = wait_event_interruptible(priv->wait,
!list_empty(used) || !priv->enabled);
> + if (ret)
> + return ret;
> +
> + spin_lock_irq(&priv->lock);
> + }
> +
if (!priv->enabled) {
spin_unlock_irq(&priv->lock);
return -ENXIO;
}
if (reader->buf) {
dbuf = reader->buf;
} else {
dbuf = list_first_entry(used, struct data_buf, entry);
list_del_init(&dbuf->entry);
}
> + /* Grab the first buffer off of the used list */
> + dbuf = list_first_entry(used, struct data_buf, entry);
> + list_del_init(&dbuf->entry);
> +
> + spin_unlock_irq(&priv->lock);
> +
> + /* Buffers are always mapped: unmap it */
> + data_unmap_buffer(priv->dev, dbuf);
> +
> + /* save the buffer for later */
> + reader->buf = dbuf;
> + reader->buf_start = 0;
> +
> + /* we removed a buffer from the used list: wake any waiters */
> + wake_up(&priv->wait);
I do not think anyone waits on this...
> +
> +have_buffer:
> + /* Get the number of bytes available */
> + avail = dbuf->size - reader->buf_start;
> + data = dbuf->vb.vaddr + reader->buf_start;
> +
> + /* Get the number of bytes we can transfer */
> + count = min(count, avail);
> +
> + /* Copy the data to the userspace buffer */
> + if (copy_to_user(ubuf, data, count))
> + return -EFAULT;
> +
> + /* Update the amount of available space */
> + avail -= count;
> +
> + /* Lock against concurrent enable/disable */
> + ret = mutex_lock_interruptible(&priv->mutex);
> + if (ret)
> + return ret;
Mutex is not needed here, we shall rely on priv->lock. Map buffer
beforehand, take lock, check if the device is enabled and if it still is
put the buffer on free list. Release lock and exit if device was
enabled; otherwise dispose the buffer.
> +
> + /* Still some space available: save the buffer for later */
> + if (avail != 0) {
> + reader->buf_start += count;
> + reader->buf = dbuf;
> + goto out_unlock;
> + }
> +
> + /*
> + * No space is available in this buffer
> + *
> + * This is a complicated decision:
> + * - if the device is not enabled: free the buffer
> + * - if the buffer is too small: free the buffer
> + */
> + if (!priv->enabled || dbuf->size != priv->bufsize) {
> + data_free_buffer(dbuf);
> + reader->buf = NULL;
> + goto out_unlock;
> + }
> +
> + /*
> + * The buffer is safe to recycle: remap it and finish
> + *
> + * If this fails, we pretend that the read never happened, and return
> + * -EFAULT to userspace. They'll retry the read again.
> + */
> + ret = data_map_buffer(priv->dev, dbuf);
> + if (ret) {
> + dev_err(priv->dev, "unable to remap buffer for DMA\n");
> + count = -EFAULT;
> + goto out_unlock;
> + }
> +
> + /* Add the buffer back to the free list */
> + reader->buf = NULL;
> + spin_lock_irq(&priv->lock);
> + list_add_tail(&dbuf->entry, &priv->free);
> + spin_unlock_irq(&priv->lock);
> +
> +out_unlock:
> + mutex_unlock(&priv->mutex);
> + return count;
> +}
> +
> +static unsigned int data_poll(struct file *filp, struct poll_table_struct *tbl)
> +{
> + struct fpga_reader *reader = filp->private_data;
> + struct fpga_device *priv = reader->priv;
> + unsigned int mask = 0;
> +
> + poll_wait(filp, &priv->wait, tbl);
> +
> + spin_lock_irq(&priv->lock);
> +
Like I mentioned, the spinlock is not useful here - all threads will get
woken up and will try to read since you are not resetting the wakeup
condition.
> + if (!list_empty(&priv->used))
> + mask |= POLLIN | POLLRDNORM;
I think you should also add:
if (!priv->)
mask |= POLLHUP | POLLERR;
to tell the readers that they should close their file descriptors.
Thanks.
--
Dmitry
^ permalink raw reply
* Re: [PATCH] sched: provide scheduler_ipi() callback in response to smp_send_reschedule()
From: Benjamin Herrenschmidt @ 2011-02-09 6:14 UTC (permalink / raw)
To: Peter Zijlstra
Cc: linux-m32r-ja, linux-mips, linux-ia64, linux-sh, H. Peter Anvin,
Heiko Carstens, Paul Mackerras, Helge Deller, sparclinux,
Linux-Arch, linux-s390, Jesper Nilsson, Jeremy Fitzhardinge,
Russell King, Hirokazu Takata, x86, James E.J. Bottomley,
virtualization, Ingo Molnar, Matt Turner, Fenghua Yu,
Mike Frysinger, user-mode-linux-devel, Konrad Rzeszutek Wilk,
Jeff Dike, Chris Metcalf, xen-devel, Mikael Starvik, linux-m32r,
Ivan Kokshaysky, user-mode-linux-user, uclinux-dist-devel,
Thomas Gleixner, linux-arm-kernel, Richard Henderson, Tony Luck,
linux-parisc, linux-cris-kernel, linux-am33-list, linux-kernel,
Ralf Baechle, Kyle McMartin, Paul Mundt, linux-alpha,
Martin Schwidefsky, linux390, Koichi Yasutake, linuxppc-dev,
David S. Miller
In-Reply-To: <1297086859.13327.16.camel@laptop>
On Mon, 2011-02-07 at 14:54 +0100, Peter Zijlstra wrote:
> On Mon, 2011-02-07 at 10:26 +1100, Benjamin Herrenschmidt wrote:
> > You missed:
> >
> > diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
> > index 9813605..467d122 100644
> > --- a/arch/powerpc/kernel/smp.c
> > +++ b/arch/powerpc/kernel/smp.c
> > @@ -98,6 +98,7 @@ void smp_message_recv(int msg)
> > break;
> > case PPC_MSG_RESCHEDULE:
> > /* we notice need_resched on exit */
> > + scheduler_ipi();
> > break;
> > case PPC_MSG_CALL_FUNC_SINGLE:
> > generic_smp_call_function_single_interrupt();
> >
> > Fold that in and add:
> >
> > Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
>
> Thanks Ben!
BTW. I we lurking at some of our CPU hotplug code and I think I cannot
totally guarantee that this won't be called on an offline CPU. If that's
a problem, you may want to add a test for that.
IE. The call function IPIs are normally not going to be sent to an
offlined CPU, and stop_machine should be a good enough fence here, but
we do abuse reschedule for a number of things (including in some case
to wake up a sleeping CPU that was pseudo-offlined :-)
Cheers,
Ben.
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox