First of all, i am sorry that i could not get the exact point of your question.
So, i will give my thought for two cases.
1) Do the memory mapped I/O for a pass-thru region of AMCC from the kernel space
including RT-tasks.
After converting the physical address into the virtual address in the kernel space
using ioremap(),
the kernel device driver or the RT-tasks can access the region.
1) Do the memory mapped I/O for a pass-thru region of AMCC from the user space.
In this case, i think the code you attached should work well.
Or, the application program in the user space can get the physical address and
size of
the pass-thru region by a IOCTL call which is processed by the kernel device
driver,
and then just remap it to the user space using the mmap function
with the opened file descriptor of "/dev/mem".
Bye.
----- Original Message -----
From: "Jens Lauer" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Tuesday, May 09, 2000 8:06 PM
Subject: [rtl] convert driver from NonRT to RT
> Hi RTer,
>
> I have a problem regarding converting a drivers code from NonRT to RT
> (RTL/RTAI). The application I have will not have any value if not
> working in realtime so I have no choice ....
>
> The driver is for a amcc PCI controller and maps one (or up to 4) pass
> thru regions of the amcc provided memory on the pci addon board to the
> user application following this scheme:
>
> a) in the drivers code there will the processing of the mmap prepared as
> follow:
>
> static void
> amcc_s5933_pt_setup (struct amcc_s5933 *dev, u32 badr, int number)
> {
> u32 mask;
>
> /* we only support memory mapped PT regions */
>
> if (badr & PCI_BASE_ADDRESS_SPACE_IO)
> {
> pr_debug ("amcc: pt: PassThru region %d IO-mapped", number);
> return;
> }
> else
> {
> pr_debug ("amcc: pt: PassThru region %d mem-mapped", number);
> }
>
> /* get configuration */
>
> cli ();
> /*
> * writing all 1 to PCI_BASE_ADDR_x
> * reading the in this register set value (set by Controller)
> * writing back the old value from init
> */
> pcibios_write_config_dword (dev->bus, dev->function, badrs[number],
> ~0);
> pcibios_read_config_dword (dev->bus, dev->function, badrs[number],
> &mask);
> pcibios_write_config_dword (dev->bus, dev->function, badrs[number],
> badr);
> sti ();
>
>
> pr_debug ("amcc: pt: %d - mask %08lx, currValue %08lx\n", number,
> mask, badr);
>
> /* extract data
> * in this way we get the size(-1) of the PCI region
> */
>
> mask &= PCI_BASE_ADDRESS_MEM_MASK;
>
> dev->ptnum++;
>
> (dev->pts[number - 1]).size = (~mask) + 1;
> (dev->pts[number - 1]).phys_addr = (badr &
> PCI_BASE_ADDRESS_MEM_MASK);
> (dev->pts[number - 1]).virt_addr = ioremap ((dev->pts[number -
> 1]).phys_addr,
> (dev->pts[number -
> 1]).size);
>
> pr_debug ("amcc: PassThru: PTNo[%d] - Size: %d, PhysAddr: %08lx,
> VirtAddr: %08lx\n", (number - 1),
> (dev->pts[number - 1]).size,
> (dev->pts[number - 1]).phys_addr,
> (dev->pts[number - 1]).virt_addr);
> }
>
>
>
> b) then the mmap is part of struct file_ops
> static int amcc_s5933_mmap ( struct file *file, struct vm_area_struct
> *vma )
> {
>
> int AREA = 0;
> unsigned long offset = vma->vm_offset;
>
> /* very simple for the moment - only pts[AREA] */
>
> switch(MINOR(file->f_dentry->d_inode->i_rdev))
> {
> case
> AMCC_PT1_MMAP_MINOR:
> {
> pr_debug("amcc: MINOR 1 -- first PT Area\n");
> AREA = 0;
> }
> break;
> case AMCC_PT2_MMAP_MINOR:
> {
> pr_debug("amcc: MINOR 2 -- second PT Area\n");
> AREA = 1;
> }
> break;
> case
> AMCC_PT3_MMAP_MINOR:
> {
> pr_debug("amcc: MINOR 3 -- third PT Area\n");
> AREA = 3;
> }
> break;
> case AMCC_PT4_MMAP_MINOR:
> {
> pr_debug("amcc: MINOR 4 -- forth PT Area\n");
> AREA = 4;
> }
> break;
>
> } file://switch
>
>
> pr_debug ("amcc: mmap: start: %08lx, end: %08lx, offset: %08lx,
> flags: %08lx\n",
> vma->vm_start, vma->vm_end, vma->vm_offset,
> vma->vm_flags);
>
> /* check dimensions */
>
> if ((vma->vm_end - vma->vm_start + vma->vm_offset) >
> amcc_dev.pts[AREA].size)
> return -EINVAL;
>
>
> /* get real address */
>
> vma->vm_offset += amcc_dev.pts[AREA].phys_addr;
> if (vma->vm_offset & ~PAGE_MASK)
> return -ENXIO;
>
> /* map pages */
>
>
>
> if (offset >= __pa(high_memory) || (file->f_flags & O_SYNC))
> {
> pgprot_val(vma->vm_page_prot) =
> pgprot_noncached(pgprot_val(vma->vm_page_prot));
> vma->vm_flags |= VM_IO;
> pr_debug("pgprot_noncached ...");
> }
>
>
> if ((vma->vm_offset) & (PAGE_SIZE-1))
> {
> pr_debug("amcc: pages not aligned\n");
> return -ENXIO;
> }
>
> if ((vma->vm_end - vma->vm_start) > (0x20000 - vma->vm_offset))
> {
> pr_debug("amcc: PT too big !!!\n");
> return -EINVAL;
> }
>
>
> // remapping | virtuell address | physical address | size |
> protection */
> if (remap_page_range (vma->vm_start, vma->vm_offset,(vma->vm_end -
> vma->vm_start), vma->vm_page_prot))
> {
> return -EAGAIN;
> }
> else
> {
> pr_debug("vma->vm_start(virtaddr): %08lx
> vma->vm_offset(physaddr):
> %08lx\n(vma->vm_end - vma->vm_start)(size): %08lx
> vma->vm_page_prot(protection): %08lx\n)",
> vma->vm_start, vma->vm_offset,(vma->vm_end - vma->vm_start),
> vma->vm_page_prot);
> }
>
> /*
> vma->vm_inode = inode;
> inode->i_count++;
> */
>
> return 0;
> }
>
> c) the app uses the mmap in this manner:
>
> pt = (__u32 *) mmap (Start, 0x20000, (PROT_READ | PROT_WRITE),
> MAP_SHARED,
> amcc_fd, (off_t) 0L); // 128KByte mapped
>
>
> Now the question: if I will have to do the handling of the memory
> formerly mapped in user spaced in my module now, what will I have to do?
> I am not the original autor of the modules code and have only some
> overview.
>
>
>
>
> --
> Jens Lauer
> TU Chemnitz /Germany
> 09107 Chemnitz
> Fakultaet fuer Elektrotechnik und Informationstechnik
> Lehrstuhl Elektrische Maschinen und Antriebe (Prof. Hofmann)
>
> Tel +49 (0)371 531 3379
> Fax +49 (0)371 531 3324
> [EMAIL PROTECTED]
> -- [rtl] ---
> To unsubscribe:
> echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
> echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
> ---
> For more information on Real-Time Linux see:
> http://www.rtlinux.org/rtlinux/
>
-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
---
For more information on Real-Time Linux see:
http://www.rtlinux.org/rtlinux/
