* Strange PCI mmap problem on MPC8555
@ 2007-07-13 11:21 willy jacobs
0 siblings, 0 replies; only message in thread
From: willy jacobs @ 2007-07-13 11:21 UTC (permalink / raw)
To: linuxppc-embedded
Setup: ELDK 4.1, U-Boot 1.2.0 and Linux 2.6.20.2 (arch/ppc)
The PPC can communicate (memory mapped) with a FGPA through the PCI bus.
A user program should be able to access (read/write) the FPGA memory areas
directly via mmap() through a small Linux driver.
In the driver initialisation bar0 (1 KByte memory) the PCI physical
address
is remapped to virtual address space via ioremap().
A memory dump (with expected values) in the kernel driver shows:
0 1 2 3 4 5 6 7 8 9 A B C D E F
0xe1056000 17 31 04 00 02 00 00 00 07 20 07 03 00 00 00 00
The mmap call in driver looks like this:
static int
rvc_mmap (struct file *filp, struct vm_area_struct *vma)
{
...
if (remap_pfn_range(vma,
vma->vm_start,
phys_bar0 >> PAGE_SHIFT,
vma->vm_end - vma->vm_start,
vma->vm_page_prot)) {
printk(KERN_CRIT "rvc_mmap: remap_page bar0 failed\n");
return -EAGAIN;
...
}
The user space test program does a mmap() to bar0 and the memory dump
(erroneous values) now shows:
0 1 2 3 4 5 6 7 8 9 A B C D E F
0x30019000 17 31 04 00 17 31 04 00 07 20 07 03 07 20 07 03
^^^^^^^^^^^ ^^^^^^^^^^^
So the first 32-bit word is now exactly the same as the second 32-bit
word,
the 4th 32-bit word is the same as the 3rd 32-bit word, etc.
Also the same behaviour for bar1.
It looks like a 32-/64 bit problem, but the kernel I use is 32-bits.
Any idea what causes this strange behaviour?
--
willy
Unclassified
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2007-07-13 13:52 UTC | newest]
Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2007-07-13 11:21 Strange PCI mmap problem on MPC8555 willy jacobs
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.