Skip to content

S
Simple PCIe FMC carrier SPEC
Commit f8a9221e authored by Federico Vaga's avatar Federico Vaga
Browse files
Options

sw:drv: add DMA read interface in debugfs 

Signed-off-by: Federico Vaga's avatarFederico Vaga <federico.vaga@cern.ch>
parent 57176098
Hide whitespace changes
Inline Side-by-side
Showing with 183 additions and 8 deletions
pytest/test_dma.py
View file @ f8a9221e
......@@ -3,6 +3,7 @@ SPDX-License-Identifier: GPL-3.0-or-later
SPDX-FileCopyrightText: 2020 CERN
"""
import pytest
import random
from PySPEC import PySPEC
class TestDma(object):
......@@ -22,3 +23,15 @@ class TestDma(object):
with pytest.raises(OSError) as error:
spec_c.dma_start()
spec.dma_stop()
@pytest.mark.parametrize("dma_size",
[random.randrange(0, 1 * 1024 * 1024, 4) for i in range(100)])
def test_dma_read(self, spec, dma_size):
"""
Read from SPEC DDR at random offsets. We just want to see if the DMA
engine reports errors
"""
spec.dma_start()
data = spec.dma_read(dma_size)
assert len(data) == dma_size
spec.dma_stop()
software/PySPEC/PySPEC/PySPEC.py
View file @ f8a9221e
......@@ -8,6 +8,8 @@ SPDX-FileCopyrightText: 2020 CERN (home.cern)
import os
class PySPEC:
DDR_SIZE = 256 * 1024 * 1024
def __init__(self, pci_id):
self.pci_id = pci_id
self.debugfs = "/sys/kernel/debug/0000:{:s}".format(self.pci_id)
......@@ -15,8 +17,14 @@ class PySPEC:
self.dma = os.path.join(self.debugfs_fpga, "dma")
def dma_start(self):
self.dma_file = open(self.dma, "r")
self.dma_file = open(self.dma, "rb")
def dma_stop(self):
if hasattr(self, "dma_file"):
self.dma_file.close()
def dma_read(self, size):
"""
Trigger a *device to memory* DMA transfer
"""
return self.dma_file.read(size)
software/kernel/spec-core-fpga.c
View file @ f8a9221e
......@@ -16,7 +16,10 @@
#include <linux/bitops.h>
#include <linux/fmc.h>
#include <linux/delay.h>
#include <linux/completion.h>
#include <linux/jiffies.h>
#include "linux/printk.h"
#include "spec.h"
#include "spec-compat.h"
......@@ -125,7 +128,8 @@ static const struct file_operations spec_fpga_dbg_bld_ops = {
struct spec_fpga_dbg_dma {
struct spec_fpga *spec_fpga;
struct dma_chan *dchan;
struct dmaengine_result dma_res;
struct completion compl;
};
static loff_t spec_fpga_dbg_dma_llseek(struct file *file,
loff_t pos, int whence)
......@@ -147,18 +151,167 @@ static loff_t spec_fpga_dbg_dma_llseek(struct file *file,
return -ENODEV;
}
static ssize_t spec_fpga_dbg_dma_read(struct file *file,
char __user *buf,
struct spec_fmca_dbg_dma_tx_ctxt {
struct dmaengine_result dma_res;
};
static void spec_fmca_dbg_dma_tx_complete(void *arg,
const struct dmaengine_result *result)
{
struct spec_fpga_dbg_dma *dbgdma = arg;
memcpy(&dbgdma->dma_res, result, sizeof(*result));
complete(&dbgdma->compl);
}
static int spec_fpga_dbg_dma_transfer(struct spec_fpga_dbg_dma *dbgdma,
enum dma_transfer_direction dir,
void *data, size_t count, loff_t offset)
{
size_t max_segment_size;
unsigned int nr_pages = (count / PAGE_SIZE) + 1;
struct page **pages;
struct sg_table sgt;
int i;
int err;
int sg_len;
struct dma_slave_config sconfig;
struct dma_async_tx_descriptor *tx;
dma_cookie_t cookie;
pages = kcalloc(nr_pages, sizeof(struct page *), GFP_KERNEL);
if (!pages)
return -ENOMEM;
for (i = 0; i < nr_pages; ++i)
pages[i] = virt_to_page(data + PAGE_SIZE * i);
max_segment_size = dma_get_max_seg_size(dbgdma->dchan->device->dev);
max_segment_size &= PAGE_MASK; /* to make alloc_table happy */
err = __sg_alloc_table_from_pages(&sgt, pages, nr_pages,
offset_in_page(data), count,
max_segment_size, GFP_KERNEL);
if (err)
goto err_alloc;
sg_len = dma_map_sg(dbgdma->spec_fpga->dev.parent,
sgt.sgl, sgt.nents, dir);
if (sg_len < 0) {
err = sg_len;
goto err_map;
}
memset(&sconfig, 0, sizeof(sconfig));
sconfig.direction = dir;
sconfig.src_addr = offset;
err = dmaengine_slave_config(dbgdma->dchan, &sconfig);
if (err)
goto err_cfg;
tx = dmaengine_prep_slave_sg(dbgdma->dchan, sgt.sgl, sg_len,
DMA_DEV_TO_MEM, 0);
if (!tx) {
err = -EINVAL;
goto err_prep;
}
/* Setup the DMA completion callback */
dbgdma->dma_res.result = DMA_TRANS_NOERROR;
dbgdma->dma_res.residue = 0;
tx->callback_result = spec_fmca_dbg_dma_tx_complete;
tx->callback_param = (void *)dbgdma;
cookie = dmaengine_submit(tx);
if (cookie < 0) {
err = cookie;
goto err_sub;
}
dma_async_issue_pending(dbgdma->dchan);
err = wait_for_completion_interruptible_timeout(
&dbgdma->compl, msecs_to_jiffies(60000));
if (err == 0)
err = -ETIMEDOUT;
if (err > 0) {
switch (dbgdma->dma_res.result) {
case DMA_TRANS_NOERROR:
err = 0;
break;
default:
err = -EIO;
break;
}
}
err_sub:
err_prep:
err_cfg:
dma_unmap_sg(dbgdma->spec_fpga->dev.parent,
sgt.sgl, sgt.nents, dir);
err_map:
sg_free_table(&sgt);
err_alloc:
kfree(pages);
return err;
}
static ssize_t spec_fpga_dbg_dma_read(struct file *file, char __user *buf,
size_t count, loff_t *ppos)
{
return -ENODEV;
int err;
void *data;
count = min(KMALLOC_MAX_SIZE, count);
data = kmalloc(count, GFP_KERNEL);
if (!data)
return -ENOMEM;
err = spec_fpga_dbg_dma_transfer(file->private_data, DMA_DEV_TO_MEM,
data, count, *ppos);
if (err)
goto err_trans;
err = copy_to_user(data, buf, count);
if (err)
goto err_cpy;
kfree(data);
*ppos += count;
return count;
err_cpy:
err_trans:
kfree(data);
return err;
}
static ssize_t spec_fpga_dbg_dma_write(struct file *file,
const char __user *buf,
size_t count, loff_t *ppos)
const char __user *buf, size_t count,
loff_t *ppos)
{
return -ENODEV;
int err;
void *data;
count = max(KMALLOC_MAX_SIZE, count);
data = kmalloc(count, GFP_KERNEL);
if (!data)
return -ENOMEM;
err = copy_from_user(data, buf, count);
if (err)
goto err_cpy;
err = spec_fpga_dbg_dma_transfer(file->private_data, DMA_MEM_TO_DEV,
data, count, *ppos);
if (err)
goto err_trans;
kfree(data);
*ppos += count;
return count;
err_trans:
err_cpy:
kfree(data);
return err;
}
static bool spec_fpga_dbg_dma_filter(struct dma_chan *dchan, void *arg)
......@@ -195,6 +348,7 @@ static int spec_fpga_dbg_dma_open(struct inode *inode, struct file *file)
err = -EBUSY;
goto err_req;
}
init_completion(&dbgdma->compl);
dbgdma->spec_fpga = spec_fpga;
file->private_data = dbgdma;
return 0;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment