Re: question about mapping RAM into user space.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi,

I use remap_pfn_range to do mmap, and it can work.

Best Regards,
Peter Chen


On Thu, 2009-03-26 at 14:03 +0800, yihect wrote:
It seems the mux_vma_nopage() operation hadn't been called.
----- Original Message -----
From: yihect
To: kernelnewbie
Cc: yihect@xxxxxxx
Sent: Thursday, March 26, 2009 1:48 PM
Subject: question about mapping RAM into user space.


Hi, all:
    I'm a newbie of kernel. I'm working on my mux driver in hand, and want to map some RAM(say, 1024*1024*2 bytes totally)into user space.
 
    Accroding the LDD3 book, I'm using __get_free_pages() which be called in nopage() operation of struct vm_operations_struct like follow:
 
    /* get page from mux dev, and return it to the user. */
 struct page *mux_vma_nopage(struct vm_area_struct *vma,
 unsigned long address, int *type)
{
    unsigned long offset;
    struct mux_dev *dev = vma->vm_private_data;
    struct page *page = NOPAGE_SIGBUS;
    void *pageptr = NULL; /* default to "missing" */
 

    printk("In mux_vma_nopage ############################################\n" );
 
    down(&dev->sem);
    offset = (address - vma->vm_start) + (vma->vm_pgoff << PAGE_SHIFT);
    if (offset >= dev->size) goto out; /* out of range */
 
    /* get page from our dev, get free page if inavialable */
    offset >>= PAGE_SHIFT; /* offset is a number of pages */
    if (!dev->data[offset])
    {
     dev->data[offset] = (void *)__get_free_pages(GFP_KERNEL, 0); /* order must be 0 */
     if (!dev->data[offset])
         goto out;
     memset(dev->data[offset], 0, PAGE_SIZE);
    }
    pageptr = dev->data[offset];
    page = virt_to_page(pageptr);
 
    /* now increment the count */
    get_page(page);
    if (type)
         *type = VM_FAULT_MINOR;

out:
    up(&dev->sem);
    return page;
}
 
 
    The mmap() operation of fops is like:
 
int nc_mux_mmap(struct file *filp, struct vm_area_struct *vma)
{
    /* don't do anything here: "nopage" will set up page table entries */
    vma->vm_ops = &mux_vm_ops;
    vma->vm_flags |= VM_RESERVED;
    vma->vm_private_data = filp->private_data;
    mux_vma_open(vma);
    return 0;
}
 
     When I test useing code :
address=mmap(0, len, PROT_READ, MAP_FILE | MAP_PRIVATE, fd, offset);
 
    if (address == (void *)-1) {
        fprintf(stderr,"%s: mmap(): %s\n",argv[0],strerror(errno));
        exit(1);
    }
    strcpy((char *)address, "hello, world.");

It gives me msg of "segmen fault":
 
/mnt/nfs $ ./mapper /dev/mux_dev 0x0 0x400
SDVR22xxMX MUX Device Driver Open...
pgd = c01a4000
[40018000] *pgd=001e8031, *pte=00000000, *ppte=00000000
 
Pid: 151, comm:               mapper
CPU: 0
PC is at 0x4008b9f8
LR is at 0x8960
pc : [<4008b9f8>]    lr : [<00008960>]    Tainted: P    
sp : bec9a64c  ip : 4008b9ec  fp : bec9ae88
r10: 40133000  r9 : 00000000  r8 : 00008738
r7 : 00000004  r6 : 00008b3c  r5 : 00008ad8  r4 : bec9aeb4
r3 : 00000068  r2 : 4000f3c3  r1 : 00008c3d  r0 : 40018000
Flags: nZCv  IRQs on  FIQs on  Mode USER_32  Segment user
Control: 397F  Table: 001A4000  DAC: 00000015
[<c052e7d8>] (show_regs+0x0/0x4c) from [<c0533bcc>] (__do_user_fault+0x5c/0xa4)
 r4 = C0039820
[<c0533b70>] (__do_user_fault+0x0/0xa4) from [<c0533e98>] (do_page_fault+0x218/0x250)
 r7 = C05240CC  r6 = C0047FB0  r5 = C0039820  r4 = FFFFFFEB
[<c0533c80>] (do_page_fault+0x0/0x250) from [<c053401c>] (do_DataAbort+0x3c/0xa0)
[<c0533fe0>] (do_DataAbort+0x0/0xa0) from [<c052c908>] (ret_from_exception+0x0/0x10)
 r8 = 00008738  r7 = 00000004  r6 = 00008B3C  r5 = 00008AD8
 r4 = FFFFFFFF
 SDVR22xxMX MUX Device Driver Release...
Segmentation fault
 
Platform is arm and kernel version is 2.6.14, I don'e know how to do nextly, is somebody have any suggestion? thanks.

[Index of Archives]     [Newbies FAQ]     [Linux Kernel Mentors]     [Linux Kernel Development]     [IETF Annouce]     [Git]     [Networking]     [Security]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux RAID]     [Linux SCSI]     [Linux ACPI]
  Powered by Linux