search for: gk20a_ram_put

Fix a very shameful memory leak and a compilation error due to the use of non-exported CMA functions. The workaround for the latter is not really elegant (replace the CMA functions by a runtime failure if we are compiled as a module), but is temporary and still an improvement over the current situation (compile error). Alexandre Courbot (2): drm/gk20a/fb: fix huge memory leak drm/gk20a/fb:

gk20a_ram_put() can be called with a NULL nouveau_mem in case of error. Handle that case the way is it done in other RAM drivers. Signed-off-by: Alexandre Courbot <acourbot at nvidia.com> --- drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/g...

On Mon, Apr 21, 2014 at 03:02:16PM +0900, Alexandre Courbot wrote: [...] > diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c b/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c [...] > + pages = dma_alloc_from_contiguous(dev, ncmin, order); > + if (!pages) { > + gk20a_ram_put(pfb, &mem); > + return -ENOMEM; > + } > + > + dma_addr = pfn_to_dma(nv_device_base(nv_device(pfb)), > + page_to_pfn(pages)); This breaks compilation on x86 because neither pfn_to_dma() nor dma_to_pfn() are available. Is there some other way this can be allocated so...

.../nouveau/core/subdev/fb/ramgk20a.c >>> b/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c >> >> [...] >>> >>> + pages = dma_alloc_from_contiguous(dev, ncmin, order); >>> + if (!pages) { >>> + gk20a_ram_put(pfb, &mem); >>> + return -ENOMEM; >>> + } >>> + >>> + dma_addr = pfn_to_dma(nv_device_base(nv_device(pfb)), >>> + page_to_pfn(pages)); >> >> >> Th...

...t;linux/types.h> -#include <linux/dma-contiguous.h> +#include <linux/mm.h> +#include <linux/dma-mapping.h> + +struct gk20a_mem { + struct nouveau_mem base; + void *cpuaddr; + dma_addr_t handle; +}; +#define to_gk20a_mem(m) container_of(m, struct gk20a_mem, base) static void gk20a_ram_put(struct nouveau_fb *pfb, struct nouveau_mem **pmem) { struct device *dev = nv_device_base(nv_device(pfb)); - struct nouveau_mem *mem = *pmem; - int i; + struct gk20a_mem *mem = to_gk20a_mem(*pmem); *pmem = NULL; if (unlikely(mem == NULL)) return; - for (i = 0; i < mem->size; i++)...

..., ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ + +#include "priv.h" + +#include <subdev/fb.h> + +#include <linux/mm.h> +#include <linux/types.h> +#include <linux/dma-contiguous.h> + +static void +gk20a_ram_put(struct nouveau_fb *pfb, struct nouveau_mem **pmem) +{ + struct device *dev = nv_device_base(nv_device(pfb)); + struct nouveau_mem *mem = *pmem; + int i; + + *pmem = NULL; + + for (i = 0; i < mem->size; i++) { + struct page *page; + + if (mem->pages[i] == 0) + break; + + page = pfn_to_...

...ouveau/nvkm/subdev/fb/ramgk20a.c +++ b/drm/nouveau/nvkm/subdev/fb/ramgk20a.c @@ -23,99 +23,17 @@ #include <core/device.h> -struct gk20a_mem { - struct nvkm_mem base; - void *cpuaddr; - dma_addr_t handle; -}; -#define to_gk20a_mem(m) container_of(m, struct gk20a_mem, base) - static void gk20a_ram_put(struct nvkm_fb *pfb, struct nvkm_mem **pmem) { - struct device *dev = nv_device_base(nv_device(pfb)); - struct gk20a_mem *mem = to_gk20a_mem(*pmem); - - *pmem = NULL; - if (unlikely(mem == NULL)) - return; - - if (likely(mem->cpuaddr)) - dma_free_coherent(dev, mem->base.size << PAGE_...

...14 at 03:02:16PM +0900, Alexandre Courbot wrote: > [...] >> diff --git a/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c b/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c > [...] >> + pages = dma_alloc_from_contiguous(dev, ncmin, order); >> + if (!pages) { >> + gk20a_ram_put(pfb, &mem); >> + return -ENOMEM; >> + } >> + >> + dma_addr = pfn_to_dma(nv_device_base(nv_device(pfb)), >> + page_to_pfn(pages)); > > This breaks compilation on x86 because neither pfn_to_dma() nor > dma_to_pfn() are available. Is there some o...

...> >>> b/drivers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c > >> > >> [...] > >>> > >>> + pages = dma_alloc_from_contiguous(dev, ncmin, order); > >>> + if (!pages) { > >>> + gk20a_ram_put(pfb, &mem); > >>> + return -ENOMEM; > >>> + } > >>> + > >>> + dma_addr = pfn_to_dma(nv_device_base(nv_device(pfb)), > >>> + page_to_pfn(pages)); >...

...ers/gpu/drm/nouveau/core/subdev/fb/ramgk20a.c >> >> >> >> [...] >> >>> >> >>> + pages = dma_alloc_from_contiguous(dev, ncmin, order); >> >>> + if (!pages) { >> >>> + gk20a_ram_put(pfb, &mem); >> >>> + return -ENOMEM; >> >>> + } >> >>> + >> >>> + dma_addr = pfn_to_dma(nv_device_base(nv_device(pfb)), >> >>> + page_...

Changes since v3: - Use a single dma_attr for all DMA-API allocations in instmem instead of one per allocation - Use device.info.ram_size instead of pfb->ram to check whether VRAM is present outside of nvkm Changes since v2: - Cleaner changes for ltc - Fixed typos in gk20a instmem IOMMU comments Changes since v1: - Add missing else condition in ltc - Remove extra flags that slipped into

Hi everyone, Way overdue v2 of the final patches that enable basic GK20A support. Hopefully all the issues raised with v1 have been addressed. Changes since v1: - Use gk20a clock driver by Ben instead of twiddling nv04's - Name new classes after gk20a instead of nvea - Addressed comments about BAR initialization code factorization - Removed non-essential code which only purpose was to avoid

Changes since v1: - Add missing else condition in ltc - Remove extra flags that slipped into nouveau_display.c and nv84_fence.c. Original cover letter: Patches 1-3 make the presence of a RAM device optional, and remove GK20A's dummy RAM driver we were using so far. On chips using shared memory, such a device can confuse the driver into moving objects where there is no need to, and can trick

A series I have waited too long to submit, and the recent refactoring made me pay the price of my perfectionism, so here are the features that are at least completed Patches 1-3 make the presence of a RAM device optional, and remove GK20A's dummy RAM driver we were using so far. On chips using shared memory, such a device can confuse the driver into moving objects where there is no need to,

Thanks Ilia for the v2 review! Here is the v3 of this IOMMU support for GK20A series. Changes since v2: - Cleaner changes for ltc - Fixed typos in gk20a instmem IOMMU comments Changes since v1: - Add missing else condition in ltc - Remove extra flags that slipped into nouveau_display.c and nv84_fence.c. Original cover letter: Patches 1-3 make the presence of a RAM device optional, and remove

Latest patches for GK20A, taking comments received for v3 into account. Changes since v3: - use only pfn_to_page() and page_to_pfn() in GK20A's FB. These functions are present on every arch and the physical address to page frame number conversion is also consistently a shift of PAGE_SHIFT. This part will probably be replaced by something nicer in the future anyway. - fixed a warning on

Changes since v2: - Enabled software class - Removed unneeded changes to nouveau_accel_init() - Replaced use of architecture-private pfn_to_dma() and dma_to_pfn() with the portable page_to_phys()/phys_to_page() - Fixed incorrect comment/commit log talking about bytes instead of words Hope this looks good! Once this gets merged the next set will be to use this driver on Jetson and Venice2