On Fri, Jan 07, 2011 at 12:11:44PM -0500, Konrad Rzeszutek Wilk wrote:
> If the TTM layer has used the DMA API to setup pages that are
> TTM_PAGE_FLAG_DMA32 (look at patch titled: "ttm: Utilize the dma_addr_t
> array for pages that are to in DMA32 pool."), lets use it
> when programming the GART in the PCIe type cards.
>
> This patch skips doing the pci_map_page (and pci_unmap_page) if
> there is a DMA addresses passed in for that page. If the dma_address
> is zero (or DMA_ERROR_CODE), then we continue on with our old
> behaviour.
Hey Ben and Jerome,
I should have CC-ed you guys earlier but missed that and instead just
CC-ed the mailing list. I was wondering what your thoughts are
about this patchset? Thomas took a look at the patchset and he is OK
but more eyes never hurt.
>
> Signed-off-by: Konrad Rzeszutek Wilk <konrad.w...@oracle.com>
> ---
> drivers/gpu/drm/nouveau/nouveau_sgdma.c | 28 +++++++++++++++++++++-------
> 1 files changed, 21 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/gpu/drm/nouveau/nouveau_sgdma.c
> b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
> index edc140a..bbdd982 100644
> --- a/drivers/gpu/drm/nouveau/nouveau_sgdma.c
> +++ b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
> @@ -12,6 +12,7 @@ struct nouveau_sgdma_be {
> struct drm_device *dev;
>
> dma_addr_t *pages;
> + bool *ttm_alloced;
> unsigned nr_pages;
>
> unsigned pte_start;
> @@ -35,15 +36,25 @@ nouveau_sgdma_populate(struct ttm_backend *be, unsigned
> long num_pages,
> if (!nvbe->pages)
> return -ENOMEM;
>
> + nvbe->ttm_alloced = kmalloc(sizeof(bool) * num_pages, GFP_KERNEL);
> + if (!nvbe->ttm_alloced)
> + return -ENOMEM;
> +
> nvbe->nr_pages = 0;
> while (num_pages--) {
> - nvbe->pages[nvbe->nr_pages] =
> - pci_map_page(dev->pdev, pages[nvbe->nr_pages], 0,
> + if (dma_addrs[nvbe->nr_pages] != DMA_ERROR_CODE) {
> + nvbe->pages[nvbe->nr_pages] =
> + dma_addrs[nvbe->nr_pages];
> + nvbe->ttm_alloced[nvbe->nr_pages] = true;
> + } else {
> + nvbe->pages[nvbe->nr_pages] =
> + pci_map_page(dev->pdev, pages[nvbe->nr_pages],
> 0,
> PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
> - if (pci_dma_mapping_error(dev->pdev,
> - nvbe->pages[nvbe->nr_pages])) {
> - be->func->clear(be);
> - return -EFAULT;
> + if (pci_dma_mapping_error(dev->pdev,
> + nvbe->pages[nvbe->nr_pages]))
> {
> + be->func->clear(be);
> + return -EFAULT;
> + }
> }
>
> nvbe->nr_pages++;
> @@ -66,11 +77,14 @@ nouveau_sgdma_clear(struct ttm_backend *be)
> be->func->unbind(be);
>
> while (nvbe->nr_pages--) {
> - pci_unmap_page(dev->pdev, nvbe->pages[nvbe->nr_pages],
> + if (!nvbe->ttm_alloced[nvbe->nr_pages])
> + pci_unmap_page(dev->pdev,
> nvbe->pages[nvbe->nr_pages],
> PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
> }
> kfree(nvbe->pages);
> + kfree(nvbe->ttm_alloced);
> nvbe->pages = NULL;
> + nvbe->ttm_alloced = NULL;
> nvbe->nr_pages = 0;
> }
> }
> --
> 1.7.1
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel
