Hi Bin,
On 3 February 2016 at 04:59, Mugunthan V N mugunthanvnm@ti.com wrote:
Implement scsi_init() api to probe driver model based sata devices.
Signed-off-by: Mugunthan V N mugunthanvnm@ti.com
drivers/block/disk-uclass.c | 39 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 39 insertions(+)
This patch seems odd to me. I would hope that scsi_init() would go away with driver model and it would happen as part of the controller probe. But of course we cannot probe SCSI from UCLASS_DISK. I think the uclass 'disk' is too broad to be useful.
So I am wondering whether the decision to use UCLASS_DISK instead of UCLASS_AHCI was a mistake.
Perhaps instead we should have:
UCLASS_AHCI UCLASS_SCSI UCLASS_MMC etc...
and each of these devices can have a UCLASS_BLK under them (the block device).
Possibly we could even have a dummy UCLASS_DISK device under each of the above, but I'm not sure that is useful.
What do you think?
diff --git a/drivers/block/disk-uclass.c b/drivers/block/disk-uclass.c index d665b35..4bd7b56 100644 --- a/drivers/block/disk-uclass.c +++ b/drivers/block/disk-uclass.c @@ -7,6 +7,45 @@
#include <common.h> #include <dm.h> +#include <dm/uclass-internal.h> +#include <dm/device-internal.h> +#include <scsi.h>
+int scsi_get_device(int index, struct udevice **devp) +{
struct udevice *dev;int ret;ret = uclass_find_device(UCLASS_DISK, index, &dev);if (ret || !dev) {printf("%d device not found\n", index);return ret;}ret = device_probe(dev);if (ret) {error("device probe error\n");return ret;}*devp = dev;return ret;+}
+void scsi_init(void) +{
struct udevice *dev;int ret;ret = scsi_get_device(0, &dev);if (ret || !dev) {error("scsi device not found\n");return;}scsi_scan(1);+}
UCLASS_DRIVER(disk) = { .id = UCLASS_DISK, -- 2.7.0.75.g3ee1e0f
Regards, Simon