ccio-dma.c

这个linux源代码是很全面的~基本完整了~使用c编译的~由于时间问题我没有亲自测试~但就算用来做参考资料也是非常好的

			   (unsigned long)sg_dma_address(startsg), cnt,
			   startsg->address, startsg->length
		);

		/*
		** Look for the start of a new DMA stream
		*/
		if(sg_dma_address(startsg) & PIDE_FLAG) {
			u32 pide = sg_dma_address(startsg) & ~PIDE_FLAG;
			dma_offset = (unsigned long) pide & ~IOVP_MASK;
			sg_dma_address(startsg) = 0;
			dma_sg++;
			sg_dma_address(dma_sg) = pide;
			pdirp = &(ioc->pdir_base[pide >> IOVP_SHIFT]);
			n_mappings++;
		}

		/*
		** Look for a VCONTIG chunk
		*/
		if (cnt) {
			unsigned long vaddr = (unsigned long)startsg->address;
			ASSERT(pdirp);

			/* Since multiple Vcontig blocks could make up
			** one DMA stream, *add* cnt to dma_len.
			*/
			sg_dma_len(dma_sg) += cnt;
			cnt += dma_offset;
			dma_offset=0;	/* only want offset on first chunk */
			cnt = ROUNDUP(cnt, IOVP_SIZE);
#ifdef CONFIG_PROC_FS
			ioc->msg_pages += cnt >> IOVP_SHIFT;
#endif
			do {
				ccio_io_pdir_entry(pdirp, KERNEL_SPACE, 
						   (void *)vaddr, hint);
				vaddr += IOVP_SIZE;
				cnt -= IOVP_SIZE;
				pdirp++;
			} while (cnt > 0);
		}
		startsg++;
	}
	return(n_mappings);
}

/*
** First pass is to walk the SG list and determine where the breaks are
** in the DMA stream. Allocates PDIR entries but does not fill them.
** Returns the number of DMA chunks.
**
** Doing the fill seperate from the coalescing/allocation keeps the
** code simpler. Future enhancement could make one pass through
** the sglist do both.
*/

static CCIO_INLINE int
ccio_coalesce_chunks(struct ioc *ioc, struct scatterlist *startsg, int nents)
{
	struct scatterlist *vcontig_sg;    /* VCONTIG chunk head */
	unsigned long vcontig_len;         /* len of VCONTIG chunk */
	unsigned long vcontig_end;
	struct scatterlist *dma_sg;        /* next DMA stream head */
	unsigned long dma_offset, dma_len; /* start/len of DMA stream */
	int n_mappings = 0;

	while (nents > 0) {

		/*
		** Prepare for first/next DMA stream
		*/
		dma_sg = vcontig_sg = startsg;
		dma_len = vcontig_len = vcontig_end = startsg->length;
		vcontig_end += (unsigned long) startsg->address;
		dma_offset = (unsigned long) startsg->address & ~IOVP_MASK;

		/* PARANOID: clear entries */
		sg_dma_address(startsg) = 0;
		sg_dma_len(startsg) = 0;

		/*
		** This loop terminates one iteration "early" since
		** it's always looking one "ahead".
		*/
		while(--nents > 0) {
			unsigned long startsg_end;

			startsg++;
			startsg_end = (unsigned long)startsg->address + 
				startsg->length;

			/* PARANOID: clear entries */
			sg_dma_address(startsg) = 0;
			sg_dma_len(startsg) = 0;

			/*
			** First make sure current dma stream won't
			** exceed DMA_CHUNK_SIZE if we coalesce the
			** next entry.
			*/   
			if(ROUNDUP(dma_len + dma_offset + startsg->length,
				   IOVP_SIZE) > DMA_CHUNK_SIZE)
				break;

			/*
			** Append the next transaction?
			*/
			if(vcontig_end == (unsigned long) startsg->address) {
				vcontig_len += startsg->length;
				vcontig_end += startsg->length;
				dma_len     += startsg->length;
				continue;
			}

			/*
			** Not virtually contigous.
			** Terminate prev chunk.
			** Start a new chunk.
			**
			** Once we start a new VCONTIG chunk, dma_offset
			** can't change. And we need the offset from the first
			** chunk - not the last one. Ergo Successive chunks
			** must start on page boundaries and dove tail
			** with it's predecessor.
			*/
			sg_dma_len(vcontig_sg) = vcontig_len;

			vcontig_sg = startsg;
			vcontig_len = startsg->length;
			break;
		}

		/*
		** End of DMA Stream
		** Terminate last VCONTIG block.
		** Allocate space for DMA stream.
		*/
		sg_dma_len(vcontig_sg) = vcontig_len;
		dma_len = ROUNDUP(dma_len + dma_offset, IOVP_SIZE);
		sg_dma_address(dma_sg) =
			PIDE_FLAG 
			| (ccio_alloc_range(ioc, (dma_len >> IOVP_SHIFT)) << IOVP_SHIFT)
			| dma_offset;
		n_mappings++;
	}

	return n_mappings;
}

/**
 * ccio_map_sg - Map the scatter/gather list into the IOMMU.
 * @dev: The PCI device.
 * @sglist: The scatter/gather list to be mapped in the IOMMU.
 * @nents: The number of entries in the scatter/gather list.
 * @direction: The direction of the DMA transaction (to/from device).
 *
 * This function implements the pci_map_sg function.
 */
static int
ccio_map_sg(struct pci_dev *dev, struct scatterlist *sglist, int nents, 
	    int direction)
{
	struct ioc *ioc;
	int coalesced, filled = 0;
	unsigned long flags;
	unsigned long hint = hint_lookup[direction];
	
	ASSERT(dev);
	ASSERT(dev->sysdata);
	ASSERT(HBA_DATA(dev->sysdata)->iommu);
	ioc = GET_IOC(dev);
	
	DBG_RUN_SG("%s() START %d entries\n", __FUNCTION__, nents);

	/* Fast path single entry scatterlists. */
	if(nents == 1) {
		sg_dma_address(sglist)= ccio_map_single(dev, sglist->address,
							sglist->length, 
							direction);
		sg_dma_len(sglist)= sglist->length;
		return 1;
	}
	
	spin_lock_irqsave(&ioc->res_lock, flags);

#ifdef CONFIG_PROC_FS
	ioc->msg_calls++;
#endif

	/*
	** First coalesce the chunks and allocate I/O pdir space
	**
	** If this is one DMA stream, we can properly map using the
	** correct virtual address associated with each DMA page.
	** w/o this association, we wouldn't have coherent DMA!
	** Access to the virtual address is what forces a two pass algorithm.
	*/
	coalesced = ccio_coalesce_chunks(ioc, sglist, nents);

	/*
	** Program the I/O Pdir
	**
	** map the virtual addresses to the I/O Pdir
	** o dma_address will contain the pdir index
	** o dma_len will contain the number of bytes to map 
	** o address contains the virtual address.
	*/
	filled = ccio_fill_pdir(ioc, sglist, nents, hint);

	spin_unlock_irqrestore(&ioc->res_lock, flags);

	ASSERT(coalesced == filled);
	DBG_RUN_SG("%s() DONE %d mappings\n", __FUNCTION__, filled);

	return filled;
}

/**
 * ccio_unmap_sg - Unmap the scatter/gather list from the IOMMU.
 * @dev: The PCI device.
 * @sglist: The scatter/gather list to be unmapped from the IOMMU.
 * @nents: The number of entries in the scatter/gather list.
 * @direction: The direction of the DMA transaction (to/from device).
 *
 * This function implements the pci_unmap_sg function.
 */
static void 
ccio_unmap_sg(struct pci_dev *dev, struct scatterlist *sglist, int nents, 
	      int direction)
{
	struct ioc *ioc;

	ASSERT(dev);
	ASSERT(dev->sysdata);
	ASSERT(HBA_DATA(dev->sysdata)->iommu);
	ioc = GET_IOC(dev);

	DBG_RUN_SG("%s() START %d entries,  %p,%x\n",
		__FUNCTION__, nents, sglist->address, sglist->length);

#ifdef CONFIG_PROC_FS
	ioc->usg_calls++;
#endif

	while(sg_dma_len(sglist) && nents--) {

#ifdef CONFIG_PROC_FS
		ioc->usg_pages += sg_dma_len(sglist) >> PAGE_SHIFT;
#endif
		ccio_unmap_single(dev, sg_dma_address(sglist),
				  sg_dma_len(sglist), direction);
		++sglist;
	}

	DBG_RUN_SG("%s() DONE (nents %d)\n", __FUNCTION__, nents);
}

static struct pci_dma_ops ccio_ops = {
	ccio_dma_supported,
	ccio_alloc_consistent,
	ccio_free_consistent,
	ccio_map_single,
	ccio_unmap_single,
	ccio_map_sg,
	ccio_unmap_sg,
	NULL,                   /* dma_sync_single : NOP for U2/Uturn */
	NULL,                   /* dma_sync_sg     : ditto */
};

#ifdef CONFIG_PROC_FS
static int proc_append(char *src, int len, char **dst, off_t *offset, int *max)
{
	if (len < *offset) {
		*offset -= len;
		return 0;
	}
	if (*offset > 0) {
		src += *offset;
		len -= *offset;
		*offset = 0;
	}
	if (len > *max) {
		len = *max;
	}
	memcpy(*dst, src, len);
	*dst += len;
	*max -= len;
	return (*max == 0);
}

static int ccio_proc_info(char *buf, char **start, off_t offset, int count,
			  int *eof, void *data)
{
	int max = count;
	char tmp[80]; /* width of an ANSI-standard terminal */
	struct ioc *ioc = ioc_list;

	while (ioc != NULL) {
		unsigned int total_pages = ioc->res_size << 3;
		unsigned long avg = 0, min, max;
		int j, len;

		len = sprintf(tmp, "%s\n", ioc->name);
		if (proc_append(tmp, len, &buf, &offset, &count))
			break;
		
		len = sprintf(tmp, "Cujo 2.0 bug    : %s\n",
			      (ioc->cujo20_bug ? "yes" : "no"));
		if (proc_append(tmp, len, &buf, &offset, &count))
			break;
		
		len = sprintf(tmp, "IO PDIR size    : %d bytes (%d entries)\n",
			      total_pages * 8, total_pages);
		if (proc_append(tmp, len, &buf, &offset, &count))
			break;
		
		len = sprintf(tmp, "IO PDIR entries : %ld free  %ld used (%d%%)\n",
			      total_pages - ioc->used_pages, ioc->used_pages,
			      (int)(ioc->used_pages * 100 / total_pages));
		if (proc_append(tmp, len, &buf, &offset, &count))
			break;
		
		len = sprintf(tmp, "Resource bitmap : %d bytes (%d pages)\n", 
			ioc->res_size, total_pages);
		if (proc_append(tmp, len, &buf, &offset, &count))
			break;
		
		min = max = ioc->avg_search[0];
		for(j = 0; j < CCIO_SEARCH_SAMPLE; ++j) {
			avg += ioc->avg_search[j];
			if(ioc->avg_search[j] > max) 
				max = ioc->avg_search[j];
			if(ioc->avg_search[j] < min) 
				min = ioc->avg_search[j];
		}
		avg /= CCIO_SEARCH_SAMPLE;
		len = sprintf(tmp, "  Bitmap search : %ld/%ld/%ld (min/avg/max CPU Cycles)\n",
			      min, avg, max);
		if (proc_append(tmp, len, &buf, &offset, &count))
			break;

		len = sprintf(tmp, "pci_map_single(): %8ld calls  %8ld pages (avg %d/1000)\n",
			      ioc->msingle_calls, ioc->msingle_pages,
			      (int)((ioc->msingle_pages * 1000)/ioc->msingle_calls));
		if (proc_append(tmp, len, &buf, &offset, &count))
			break;
		

		/* KLUGE - unmap_sg calls unmap_single for each mapped page */
		min = ioc->usingle_calls - ioc->usg_calls;
		max = ioc->usingle_pages - ioc->usg_pages;
		len = sprintf(tmp, "pci_unmap_single: %8ld calls  %8ld pages (avg %d/1000)\n",
			      min, max, (int)((max * 1000)/min));
		if (proc_append(tmp, len, &buf, &offset, &count))
			break;
 
		len = sprintf(tmp, "pci_map_sg()    : %8ld calls  %8ld pages (avg %d/1000)\n",
			      ioc->msg_calls, ioc->msg_pages,
			      (int)((ioc->msg_pages * 1000)/ioc->msg_calls));
		if (proc_append(tmp, len, &buf, &offset, &count))
			break;
		len = sprintf(tmp, "pci_unmap_sg()  : %8ld calls  %8ld pages (avg %d/1000)\n\n\n",
			      ioc->usg_calls, ioc->usg_pages,
			      (int)((ioc->usg_pages * 1000)/ioc->usg_calls));
		if (proc_append(tmp, len, &buf, &offset, &count))
			break;

		ioc = ioc->next;
	}

	if (count == 0) {
		*eof = 1;
	}
	return (max - count);
}

static int ccio_resource_map(char *buf, char **start, off_t offset, int len,
			     int *eof, void *data)
{
	struct ioc *ioc = ioc_list;

	buf[0] = '\0';
	while (ioc != NULL) {
		u32 *res_ptr = (u32 *)ioc->res_map;
		int j;

		for (j = 0; j < (ioc->res_size / sizeof(u32)); j++) {
			if ((j & 7) == 0)
				strcat(buf,"\n   ");
			sprintf(buf, "%s %08x", buf, *res_ptr);
			res_ptr++;
		}
		strcat(buf, "\n\n");
		ioc = ioc->next;
		break; /* XXX - remove me */
	}

	return strlen(buf);
}
#endif

/**
 * ccio_find_ioc - Find the ioc in the ioc_list