pci_alloc_consistent 配合mmap

有这种需求的同行自然明白这2个为什么需要配合起来用。简单说说，我的需求是dma位于pci设备侧，pci主的cpu上应用程序直接mmap获取dma发来的数据。

猜测显卡的dma一般位于pci主，所以我在内核里没有找到pci_alloc_consistent 配合mmap的例子。

这个代码在loongson 3A这个极品芯片上都能通过，相信其他平台毫无障碍。

pci_alloc_consistent得到的是申请到的dma一致性缓冲区的内核态虚拟地址以及pci总线地址。

内核态虚拟地址转成物理地址后需要传递给应用做mmap的最后一个参数。

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/pci.h>
#include <linux/init.h>
#include <linux/ioport.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/delay.h>
#include <linux/ethtool.h>
#include <linux/mii.h>
#include <linux/crc32.h>
#include <linux/io.h>
#include <linux/interrupt.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/types.h>
#include <linux/miscdevice.h>
#include <linux/ioport.h>
#include <linux/fcntl.h>
#include <linux/init.h>
#include <linux/poll.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/spinlock.h>
#include <linux/sched.h>
#include <linux/sysctl.h>
#include <linux/wait.h>
#include <linux/cdev.h>
#include <linux/fs.h>
#include <linux/delay.h>
#include <linux/uaccess.h>
#include <linux/device.h>
#include <linux/err.h>
#include <linux/fs.h>
#include <asm/io.h>
#include <asm/current.h>
#include <asm/system.h>
#include <linux/mm.h>
#include <linux/mman.h>
#include <linux/miscdevice.h>
#include <linux/proc_fs.h>
#include <linux/device.h>
#include <linux/fs.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/slab.h>

#define DEVICE_NAME "testc"
#define MODNAME "pmc-test"
#define MMAPBUF_LEN (16*1024*1024)

struct my_testdev{
dev_t testc_dev_num;
struct class * testc_class;
unsigned long testc_kernel_virt_addr;

struct cdev test_cdev;
unsigned int current_pointer; /*char device offset ,目前同时只能一个程序读取 */
unsigned long mmap_phyaddr;
dma_addr_t mmap_pcibus_addr;
unsigned long mmap_kvirt_addr;
};

//不得已的一个全局变量
struct my_testdev *priv;

//仅仅是为了测试程序 手头只有这个卡
static DEFINE_PCI_DEVICE_TABLE(netdrv_pci_tbl) = {
{0x8086, 0x10b9, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0 },
{0,}
};
MODULE_DEVICE_TABLE(pci, netdrv_pci_tbl);

static int testc_open(struct inode *inode, struct file *file)
{
return 0;
}

static int testc_release(struct inode *inode, struct file *file)
{

return 0;
}

static ssize_t testc_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
{

return 0;
}

static ssize_t testc_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
{

return 0;
}

static loff_t testc_llseek(struct file * file,loff_t offset,int orig)
{
return 0;
}

static int testc_mmap(struct file *file, struct vm_area_struct *vma)
{

unsigned long size = vma->vm_end - vma->vm_start;
struct my_testdev *pri=priv;
printk("inmmap ,mmap_phyaddr %lx , mmap_kvirt_addr %lx \n",pri->mmap_phyaddr,pri->mmap_kvirt_addr);
printk("start %lx end %lx off %lx \n",vma->vm_start,vma->vm_end,vma->vm_pgoff);
//pci_alloc_consistent申请到的内存已经是物理地址连续的 ，只要一个remap_pfn_range
remap_pfn_range(vma,vma->vm_start,(pri->mmap_phyaddr)>> PAGE_SHIFT,size,vma->vm_page_prot);

vma->vm_flags |= VM_RESERVED; /* avoid to swap out this VMA */
return 0;

}

static const struct file_operations testc_fops = {
.read = testc_read,
//.aio_read = generic_file_aio_read,
.write = testc_write,
//.aio_write = blkdev_aio_write,
//.fsync = blkdev_fsync,
.mmap = testc_mmap,
.open = testc_open,
.release = testc_release,
//.unlocked_ioctl = raw_ioctl,
.llseek = testc_llseek,
.owner = THIS_MODULE,
};

static int __devinit pmc_probe(struct pci_dev *pdev,const struct pci_device_id *ent)
{
int ret;
int i;
struct device *x;
struct my_testdev *pri;
printk("--------%s %d\n",__FUNCTION__,__LINE__);
pri=kmalloc(sizeof(struct my_testdev),GFP_KERNEL);
memset(pri,0,sizeof(struct my_testdev));
priv=pri;
pci_set_drvdata(pdev, pri);

//char device
if(alloc_chrdev_region(&(pri->testc_dev_num), 0, 1,DEVICE_NAME))
{
printk("err----%s %d \n",__FILE__,__LINE__);
return -1;
}
pri->testc_class=class_create(THIS_MODULE,DEVICE_NAME);
cdev_init(&(pri->test_cdev),&testc_fops);
pri->test_cdev.owner=THIS_MODULE;

ret=cdev_add(&(pri->test_cdev),pri->testc_dev_num,1);
if(ret)
{
printk("err----%s %d \n",__FILE__,__LINE__);
return -1;
}
//设备节点自动生成
x=device_create(pri->testc_class,NULL,MKDEV(MAJOR(pri->testc_dev_num),0),pri,DEVICE_NAME);
if(x==NULL)
{
printk("err----%s %d \n",__FILE__,__LINE__);
return -1;
}
priv=pri;
priv->current_pointer=0;

//mmap
pri->mmap_kvirt_addr=(unsigned long )pci_alloc_consistent(pdev,MMAPBUF_LEN,&(pri->mmap_pcibus_addr));
if((void *)pri->mmap_kvirt_addr==NULL)
{
return -1;
}
memset((void *)pri->mmap_kvirt_addr,0x0,MMAPBUF_LEN);
//写入有意义的数据，方便在应用里验证确实映射了16MB
for(i=0;i<MMAPBUF_LEN/4;i++)
{
*(((u32 *)pri->mmap_kvirt_addr)+i)=i;
}

pri->mmap_phyaddr=virt_to_phys((void *)pri->mmap_kvirt_addr);
printk("mmap_phyaddr %lx , mmap_kvirt_addr %lx \n",pri->mmap_phyaddr,pri->mmap_kvirt_addr);

return 0;
}
static void __devexit pmc_remove(struct pci_dev *pdev)
{
struct my_testdev *pri = pci_get_drvdata(pdev);
pci_free_consistent(pdev, MMAPBUF_LEN,(void *)pri->mmap_kvirt_addr, pri->mmap_pcibus_addr);

}

static struct pci_driver netdrv_pci_driver = {
.name = MODNAME,
.id_table = netdrv_pci_tbl,
.probe = pmc_probe,
.remove = __devexit_p(pmc_remove),

};

static int __init pmc_init_module(void)
{
return pci_register_driver(&netdrv_pci_driver);
}

static void __exit pmc_cleanup_module(void)
{
pci_unregister_driver(&netdrv_pci_driver);
}

module_init(pmc_init_module);
module_exit(pmc_cleanup_module);

MODULE_AUTHOR("deep_pro");
MODULE_LICENSE("GPL");
简单的应用程序，注意mmap的最后一个参数是dma缓冲区的物理地址，演示程序里根据驱动的打印写的硬编码，最终还是要靠ioctl等机制实现自动从驱动取得。

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <errno.h>
#include <limits.h>
#include <linux/kernel.h>

#include <byteswap.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/mman.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>

#include <sys/ioctl.h>
#define MMAP_LEN 16*1024*1024

int main()
{
int fd,i,ret;

char buf[0x20]={0};
unsigned int *mmap_data;

fd=open("/dev/testc",O_RDWR);
if(fd<0)
{
perror("open:");
return -1;
}
//0xf7000000 这个是缓冲区的物理地址，要从驱动里得到
mmap_data=(unsigned int *)mmap(NULL,MMAP_LEN,PROT_READ|PROT_WRITE,MAP_SHARED,fd,0xf7000000);
if(mmap_data==MAP_FAILED)
{
perror("mmap");
return -1;
}
//测试mmap后的缓冲区读写 ，先写再读,这样执行第二遍应用就能验证写入成功
for(i=14*1024*1024/4;i<(14*1024*1024+0x120)/4;i++)
{
//mmap_buf[i]=0;
printf("mmap_buf %02x :%x \n",i,*(mmap_data+i));
*(mmap_data+i)+=1;
}

munmap(mmap_data,MMAP_LEN);

close(fd);
}