Hello,
Finally I found a solution at my last problem.
Bigphysarea patch allocated memory in the KSEG0 segment (using alloc_bootmem_low_pages).
This data are accessed through the cache.
Our architecture is defined as NON_COHERENT_IO.
So I think the data could be dirty. (Am I wrong?)
I took a look in the file arch/mips/jazz/jazzdma.c, because it allocates memory with alloc_bootmem_low_pages too, and I saw they used the following functions:
virt_addr = alloc_bootmem_low_pages (...) ;
/*I replace the previous line with
bp__virt_mem = bigphysarea_alloc_pages( ...) ;
virt_addr = bp__virt_mem ;
*/
dma_cache_wback_inv(virt_addr,size);
virt_addr=KSEG1ADDR(virt_addr);
bus_addr= PHYSADDR(virt_addr);
The memory is now in the segment KSEG1, which is not accessed through a cache.
But, I notice that I have to free the bigphysarea memory with the original address or there is a PCI bus error (see the code below). (That's why I store it in bp__virt_mem value)
I don't understand why the kernel works sometimes with the address of bigphysarea (to free the memory), and sometimes deals with the address tuned by KSEG1ADDR() (before/after a DMA transfer, I wrote/read the values at this address).
Is there any explanation?
It seems too that virt_to_phys (bp__virt_mem), virt_to_bus (bp__virt_mem) are equals to PHYSADDR(virt_addr).
So I guess that virt_to_phys applies to KSEG0 memory and PHYSADDR to KSEG1. Could you tell me please if I am wrong?
And what about the __pa() macro?
Now I wonder if I can find the bigphysarea address back, because there is only an offset between bigphysarea address and the virt_addr (equals to the offset between KSEG0 and KSEG1).
What about this?
bp_virt_addr = KSEG0(PHYSADDR(virt_addr));
Thanks a lot.
NB: With an ix86 arch, the cache handling seems transparent. That's why there was no bug.
Here is A solution I wrote:
/*address used by bigphysarea. We have to store it in order to free it later with */
caddr_t bp_virt_mem = NULL;
#define NUMBER_OF_PAGES 5
/*returns the virtual kernel address you have to deal with for DMA transfers. Ie, you write the data you want to transfer in.
bus_addr = address to pass to the DMA controller*/
void * allocate_continuous_memory(void * bus_addr ){
void * virt_addr = NULL;
/* bp_virt_mem will be page aligned*/
if ( ! (bp_virt_mem = bigphysarea_alloc_pages (NUMBER_OF_PAGES, 1, GFP_KERNEL) ) )
BUG();
virt_addr= (void *) bp_virt_mem;
/*this function makes caches and coherent by writing the content of the caches back to memory, if necessary.
It also invalidates the affected part of the caches as necessary before DMA transfers from outside of memory*/
dma_cache_wback_inv(virt_addr,size);
virt_addr=KSEG1ADDR(virt_addr);
bus_addr= PHYSADDR(virt_addr);
memset(virt_addr,0, NUMBER_OF_PAGES * PAGE_SIZE);
return (virt_addr);
}
/*that's why we have to store the address returned by bigphysarea*/
void free_continuous_memory(){
if (bp__virt_mem) {
bigphysarea_free_pages (bp__virt_mem);
bp__virt_mem = NULL;
}
}
/*if you want to remap this buffer in userland*/
int mmap_lookup (struct file *file, struct vm_area_struct *vma){
unsigned long off = vma -> vm_pgoff << PAGE_SHIFT;
unsigned long vma_size = vma-> vm_end - vma->vm_start;
/*we have to map with the address returned by bigphysarea, or it will be a PCI Bus error...*/
remap_page_range (vma->vm_start, virt_to_phys (bp__virt_mem), vma- >vm_end-vma->vm_start, vma-> vm_page_prot);
return 0;
}
-----Ursprüngliche Nachricht-----
Von: linux-mips-bounce@xxxxxxxxxxxxxx [mailto:linux-mips-bounce@xxxxxxxxxxxxxx] Im Auftrag von Francois Beauregard
Gesendet: Mittwoch, 14. Juni 2006 17:23
An: linux-mips@xxxxxxxxxxxxxx
Betreff: Bigphysarea and MIPS
Hello everybody,
it's my first post on the list, so be indulgent, I'm a newbee.
I'm writing a driver using a DMA transfer between on board RAM of a card and a memory ring buffer in kernel space. The processor is a Toshiba Tx49. The PCI controller is a PLX PCI9054 rev AC. It's seems to work fine.
The ring buffer in kernel space is then mmap()'ed in user's space using remap_page_range().
The kernel is a 2.4.22 - bigphys version.
The driver writes the pattern 0123456789 in this ring buffer (as uint32_t), and transfers those data on the on board ram thanks to the DMA (pci->local).
Then I transfer those data into the same ring buffer, but 10 uint32_t further. (local->pci)
So I may read 01234567890123456789.
I have to use a continuous memory ring buffer (about 2MBytes) due to the speed of the card.
That's why I allocated it firstly with bigphysarea patch. (I will describe it later)
As I encountered a problem with this patch, I tried to use pci_alloc_consistent (arch/mips/kernel/pci-dma.c) instead (after I read the O'Reilly Linux Driver book).
With this function, the DMA transfer succeeds (the pattern is ok: 01234567890123456789), but the mmap() call raised a PCI Bus error.
(I have seen in the mailing list that I probably have to modify mk_pte_phys(). Am I wrong?)
Now, my problem with bigphysarea.
The allocation of the ring buffer is done trough bigphysarea_alloc_pages (mm/bigphysarea.c). (I add bigphysarea=[number] as described in the documentation).
After boot, sometimes I read 01234567890000000000. Sometimes 01234567890000000089.
Then the next DMA transfer returns: 01234567890000006789
I tried to pass different flags to bigphysarea_alloc_pages (GFP_ATOMIC|GFP_DMA, GFP_KERNEL|GFP_DMA ...) without success.
I took a look at the source code of the two functions, and I tried to understand the difference, but the code is too dependent on the hardware for my knowledge))-;
pci_alloc_consistent and bigphysarea patch call finally __get_free_pages, but bigphysarea uses kmalloc call whereas pci_alloc_consistent makes some operation with dma_cache_wback_inv.
I know that the flag CONFIG_NONCOHERENT_IO is set.
The addresses of the ring buffer are respectively 0x27820000 and 0x002FA000 with bigphysarea and pci_alloc_consistent.
Are'nt bigphysarea and mips compliant?
How could I solve this problem?
Did anyone succeed with bigphysarea and mips in association?
Thanks a lot.
Regards,
/*this function allocated the ring buffer,with bigphysarea or pci_alloc_consistent
* bigphysarea is a global variable
*/
int alloc_ring_buffer(struct pci_dev *dev, uint32_t number_of_pages)
{
uint32_t free_mem=0;
PRIVDATA (dev)->PC_ring_buffer_size=number_of_pages*PAGE_SIZE;
if (bigphysarea){
printk("Using Bigphysarea\n");
bigphysarea_kinfo(&Free_Mem); /*returns the size of avialable memory*/
if (free_mem<number_of_pages*PAGE_SIZE)
printk("The bigphysarea reserved memory is too short\n");
else{
unsigned long size=PRIVDATA (dev)->PC_ring_buffer_size;
/*pci_alloc_consistent seems to use the flags : GFP_ATOMIC and GFP_DMA*/
PRIVDATA (dev)->PC_ring_buffer_virt_addr= (uint32_t* )
bigphysarea_alloc_pages (number_of_pages,1,/*GFP_ATOMIC|GFP_DMA*/GFP_KERNEL);
if (PRIVDATA (dev)->PC_ring_buffer_virt_addr) {
unsigned long adr = (unsigned long)PRIVDATA (dev)->PC_ring_buffer_virt_addr;
while (size > 0) {
mem_map_reserve(virt_to_page(phys_to_virt(adr)));
adr += PAGE_SIZE;
size -= PAGE_SIZE;
}
}
PRIVDATA (dev)->PC_ring_buffer_bus_addr=virt_to_bus (PRIVDATA (dev)->PC_ring_buffer_virt_addr);
}
}else{
printk("Using pci_alloc_consistent\n");
free_mem=free_mem;/*remove warning*/
/*allocation*/
PRIVDATA (dev)->PC_ring_buffer_virt_addr= pci_alloc_consistent(dev, PRIVDATA (dev)->PC_ring_buffer_size, &PRIVDATA (dev)->PC_ring_buffer_bus_addr);
/*put a flag to all the allocated pages*/
if (PRIVDATA (dev)->PC_ring_buffer_virt_addr){
struct page *page,*end_page;
/*the last page*/
end_page=virt_to_page(PRIVDATA (dev)->PC_ring_buffer_virt_addr+PRIVDATA (dev)->PC_ring_buffer_size-1);
for (page=virt_to_page(PRIVDATA (dev)->PC_ring_buffer_virt_addr);page<=end_page;page++)
set_bit(PG_reserved,&page->flags);
}
}
if (!(PRIVDATA (dev)->PC_ring_buffer_virt_addr)){
printk("Can't allocated contiguous memory\n");
return -ENOMEM;
}
else
printk("Allocation of the contiguous memory OK\nVirtual Address: %lX\n"\
"Bus physical Address: %lX\n",(unsigned long)PRIVDATA (dev)->PC_ring_buffer_virt_addr,(unsigned long)PRIVDATA (dev)->PC_ring_buffer_bus_addr);
memset(PRIVDATA (dev)->PC_ring_buffer_virt_addr,0,number_of_pages*PAGE_SIZE);
return 0;
}
/*-----------------------------------------------------------/
* pci_alloc_consistent
*
/*-----------------------------------------------------------*/
void *pci_alloc_consistent(struct pci_dev *hwdev, size_t size,
dma_addr_t * dma_handle)
{
void *ret;
int gfp = GFP_ATOMIC;
struct pci_bus *bus = NULL;
#ifdef CONFIG_ISA
if (hwdev == NULL || hwdev->dma_mask != 0xffffffff)
gfp |= GFP_DMA;
#endif
ret = (void *) __get_free_pages(gfp, get_order(size));
if (ret != NULL) {
memset(ret, 0, size);
if (hwdev)
bus = hwdev->bus;
*dma_handle = bus_to_baddr(bus, __pa(ret));
#ifdef CONFIG_NONCOHERENT_IO
dma_cache_wback_inv((unsigned long) ret, size);
ret = UNCAC_ADDR(ret);
#endif
}
return ret;
}
/*-----------------------------------------------------------/
* bigphysarea_alloc_pages
*
/*-----------------------------------------------------------*/
caddr_t bigphysarea_alloc_pages(int count, int align, int priority)
{
range_t *range, **range_ptr, *new_range, *align_range;
caddr_t aligned_base;
if (init_level < 2)
if (init2(priority))
return 0;
new_range = NULL;
align_range = NULL;
if (align == 0)
align = PAGE_SIZE;
else
align = align * PAGE_SIZE;
/*
* Search a free block which is large enough, even with alignment.
*/
range_ptr = &free_list;
while (*range_ptr != NULL) {
range = *range_ptr;
aligned_base =
(caddr_t)((((unsigned long)range->base + align - 1) / align) * align);
if (aligned_base + count * PAGE_SIZE <=
range->base + range->size)
break;
range_ptr = &range->next;
}
if (*range_ptr == NULL)
return 0;
range = *range_ptr;
/*
* When we have to align, the pages needed for alignment can
* be put back to the free pool.
* We check here if we need a second range data structure later
* and allocate it now, so that we don't have to check for a
* failed kmalloc later.
*/
if (aligned_base - range->base + count * PAGE_SIZE < range->size) {
new_range = kmalloc(sizeof(range_t), priority);
if (new_range == NULL)
return NULL;
}
if (aligned_base != range->base) {
align_range = kmalloc(sizeof(range_t), priority);
if (align_range == NULL) {
if (new_range != NULL)
kfree(new_range);
return NULL;
}
align_range->base = range->base;
align_range->size = aligned_base - range->base;
range->base = aligned_base;
range->size -= align_range->size;
align_range->next = range;
*range_ptr = align_range;
range_ptr = &align_range->next;
}
if (new_range != NULL) {
/*
* Range is larger than needed, create a new list element for
* the used list and shrink the element in the free list.
*/
new_range->base = range->base;
new_range->size = count * PAGE_SIZE;
range->base = new_range->base + new_range->size;
range->size = range->size - new_range->size;
} else {
/*
* Range fits perfectly, remove it from free list.
*/
*range_ptr = range->next;
new_range = range;
}
/*
* Insert block into used list
*/
new_range->next = used_list;
used_list = new_range;
return new_range->base;
}
**********************************************************************
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please notify
the system manager.
This footnote confirms that this email message has been scanned for
the presence of computer viruses.
**********************************************************************
