On Wed, Feb 28, 2018 at 05:06:57PM -0600, Alan Tull wrote:

On Tue, Feb 13, 2018 at 3:24 AM, Wu Hao [off-list ref] wrote:

Hi Hao,

quoted

This patch adds compat_id support when driver creates the platform
device for dfl-fme-region. It allows dfl-fme-region platform driver
to create fpga-region with correct compat_id.

Signed-off-by: Wu Hao <redacted>
---
 drivers/fpga/dfl-fme-pr.c | 20 ++++++++++++++++++++
 drivers/fpga/dfl-fme-pr.h |  3 +++
 2 files changed, 23 insertions(+)

diff --git a/drivers/fpga/dfl-fme-pr.c b/drivers/fpga/dfl-fme-pr.c
index 526e90b..c17170b 100644
--- a/drivers/fpga/dfl-fme-pr.c
+++ b/drivers/fpga/dfl-fme-pr.c

@@ -314,6 +314,25 @@ static void fpga_fme_destroy_bridges(struct feature_platform_data *pdata)
 }

 /**
+ * fpga_fme_get_region_compat_id - read region compat_id from hardware
+ *
+ * @dev: fme device.
+ */
+static struct fpga_region_compat_id
+fpga_fme_get_region_compat_id(struct device *dev)
+{
+       struct fpga_region_compat_id compat_id;
+       void __iomem *fme_pr;
+
+       fme_pr = get_feature_ioaddr_by_id(dev, FME_FEATURE_ID_PR_MGMT);
+
+       compat_id.id_l = readq(fme_pr + FME_PR_INTFC_ID_L);
+       compat_id.id_h = readq(fme_pr + FME_PR_INTFC_ID_H);

I tried building for 32 bit ARM and ran into readq and writeq not
being defined.  v2 had proper '#indef readq/writeq', so looks like we
need that after all.

Sorry, it's missing a header file for dfl-fme-pr.h, as some files don't include
dfl.h at this version. That header file has definitions for readq/writeq.

+#include <linux/io-64-nonatomic-lo-hi.h>

will fix it.

Thanks
Hao

Alan