On 2013-05-10 17:27, H Hartley Sweeten wrote:
On Friday, May 10, 2013 6:07 AM, Ian Abbott wrote:When allocating a buffer to support asynchronous comedi commands, if a DMA coherent buffer was requested but `CONFIG_HAS_DMA` is undefined, bail out of local helper function `__comedi_buf_alloc()` with an error message. Signed-off-by: Ian Abbott <abbotti@xxxxxxxxx> --- drivers/staging/comedi/comedi_buf.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/drivers/staging/comedi/comedi_buf.c b/drivers/staging/comedi/comedi_buf.c index b6cd67a..d4be0e6 100644 --- a/drivers/staging/comedi/comedi_buf.c +++ b/drivers/staging/comedi/comedi_buf.c @@ -76,6 +76,12 @@ static void __comedi_buf_alloc(struct comedi_device *dev, struct comedi_buf_page *buf; unsigned i; + if (!IS_ENABLED(CONFIG_HAS_DMA) && s->async_dma_dir != DMA_NONE) { + dev_err(dev->class_dev, + "dma buffer allocation not supported\n"); + return; + } + async->buf_page_list = vzalloc(sizeof(*buf) * n_pages); if (async->buf_page_list) pages = vmalloc(sizeof(struct page *) * n_pages);Would it be better to put this test in comedi_buf_alloc() and return a different errno? With the check in __comedi_buf_alloc(), comedi_buf_alloc() needlessly calls __comedi_buf_free() and __comedi_buf_alloc() before returning an errno of -ENOMEM. I think an errno of -ENOSYS, or even -EINVAL, would be more appropriate.
What would a fake dma_alloc_coherent() inline stub function return if it existed? I think it might return -ENOMEM. Since __comedi_buf_alloc() is only called from comedi_buf_alloc() and could easily be changed to return an errno, I don't think it matters where the test is done.
-- -=( Ian Abbott @ MEV Ltd. E-mail: <abbotti@xxxxxxxxx> )=- -=( Tel: +44 (0)161 477 1898 FAX: +44 (0)161 718 3587 )=- _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
