On Wed, 2017-08-30 at 13:56 -0700, Dave Jiang wrote:
:
+static int pmem_attach_disk(struct device *dev,
> + struct nd_namespace_common *ndns)
> +{
> + struct pmem_device *pmem;
> + int rc;
> + struct dma_chan *chan = NULL;
> + int has_dma;
> +
> + pmem = pmem_core_setup_pmem(dev, ndns);
> + if (!pmem)
> + return -ENXIO;
> +
> + chan = dma_find_channel(DMA_MEMCPY_SG);
> + if (!chan)
> + dev_warn(dev, "Forced back to CPU, no DMA\n");
I just tested this series on a Haswell system. The above warning
messages was shown during boot-up (is this expected?), but it did not
appear to switch back to CPU. Initial IO requests to pmem (mkfs) hit
the WARN_ON_ONCE() below. pmem->sg_allocated is zero since it was not
initialized (has_dma was set to 0).
[ 897.147527] WARNING: CPU: 23 PID: 707 at
drivers/nvdimm/pmem_dma.c:171 pmem_handle_cmd_dma+0x675/0x710
It then displayed seemed-forever-repeated error messages below.
ioatdma 0000:80:04.7: ioat_timer_event: Channel halted (1)
ioatdma 0000:80:04.7: Errors:
ioatdma 0000:80:04.7: Err(0): DMA Transfer Source Address Error
ioatdma 0000:80:04.7: Reset channel...
ioatdma 0000:80:04.7: Restart channel...
ioatdma 0000:80:04.7: ioat_timer_event: Channel halted (1000)
ioatdma 0000:80:04.7: Errors:
ioatdma 0000:80:04.7: Err(12): Completion Address Error
ioatdma 0000:80:04.7: Reset channel...
ioatdma 0000:80:04.7: Restart channel...
ioatdma 0000:80:04.7: ioat_timer_event: Channel halted (1000)
:
Any thoughts? Also, how do I choose the mode between CPU and DMA?
Thanks!
-Toshi
��.n��������+%������w��{.n��������)�)�� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
