在 2010年1月30日 上午8:37，H Hartley Sweeten [off-list ref] 写道：

On Friday, January 29, 2010 5:28 PM, Andrew Morton wrote:

quoted

On Mon, 11 Jan 2010 17:05:44 +0900
Wang Qiang [off-list ref] wrote:

quoted

hi, Dear Wan

There is the patch of LCD controller driver for nuc900s.
The Linux LOGO is just fine and the FB-Test application was ok, too.

best regards
wangqiang


...

@@ -380,6 +381,48 @@ struct platform_device nuc900_device_kpi = {

     .resource       = nuc900_kpi_resource,
 };

+#ifdef CONFIG_FB_NUC900
+
+static struct resource nuc900_lcd_resource[] = {
+    [0] = {
+            .start = W90X900_PA_LCD,
+            .end   = W90X900_PA_LCD + W90X900_SZ_LCD - 1,
+            .flags = IORESOURCE_MEM,
+    },
+    [1] = {
+            .start = IRQ_LCD,
+            .end   = IRQ_LCD,
+            .flags = IORESOURCE_IRQ,
+    }
+};
+
+static u64 nuc900_device_lcd_dmamask = 0xffffffffUL;

I suspect this should have type `dma_addr_t', but `struct device' uses
open-coded u64 too.  Odd.

It makes no sense to initialise a u64 with an unsigned long value -
it's wrong on a 32-bit machine.

this:

      static u64 nuc900_device_lcd_dmamask = -1;

will work.

quoted

+struct platform_device nuc900_device_lcd = {
+    .name             = "nuc900-lcd",
+    .id               = -1,
+    .num_resources    = ARRAY_SIZE(nuc900_lcd_resource),
+    .resource         = nuc900_lcd_resource,
+    .dev              = {
+            .dma_mask               = &nuc900_device_lcd_dmamask,
+            .coherent_dma_mask      = 0xffffffffUL

And this gets initialised to 0x00000000ffffffff also.  Using -1 will fix.

Wouldn't this also work?

       .dev                    = {
               .dma_mask                       = &nuc900_device_lcd.dev.coherent_dma_mask,
               .coherent_dma_mask      = DMA_BIT_MASK(32),
       },

Then you could just get rid of nuc900_device_lcd_dmamask.

Yes, you are right.
I will use the DMA_BIT_MASK macro.
Thank you very much.

Regards,
Hartley