From: Simon Glass <sjg@chromium.org>
Date: Tue, 29 Dec 2015 12:22:49 +0000 (-0700)
Subject: dm: mmc: Try to honour the sequence order
X-Git-Tag: v2016.03-rc1~243^2~46
X-Git-Url: https://git.sur5r.net/?a=commitdiff_plain;h=4a1db6d8ab4d5e4565bda079710a028ada12ddbe;p=u-boot

dm: mmc: Try to honour the sequence order

At present we add driver-model MMC devices in the order we find them. The
'alias' order is not honoured.

It is difficult to fix this for the case where we have holes in the
sequence. But for the common case where the devices are numbered from 0
without any gaps, we can add the devices to the internal data structures
in this order.

Signed-off-by: Simon Glass <sjg@chromium.org>
---

diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c
index e6028d503f..ede5d6eeec 100644
--- a/drivers/mmc/mmc.c
+++ b/drivers/mmc/mmc.c
@@ -1780,18 +1780,28 @@ static int mmc_probe(bd_t *bis)
 #elif defined(CONFIG_DM_MMC)
 static int mmc_probe(bd_t *bis)
 {
-	int ret;
+	int ret, i;
 	struct uclass *uc;
-	struct udevice *m;
+	struct udevice *dev;
 
 	ret = uclass_get(UCLASS_MMC, &uc);
 	if (ret)
 		return ret;
 
-	uclass_foreach_dev(m, uc) {
-		ret = device_probe(m);
+	/*
+	 * Try to add them in sequence order. Really with driver model we
+	 * should allow holes, but the current MMC list does not allow that.
+	 * So if we request 0, 1, 3 we will get 0, 1, 2.
+	 */
+	for (i = 0; ; i++) {
+		ret = uclass_get_device_by_seq(UCLASS_MMC, i, &dev);
+		if (ret == -ENODEV)
+			break;
+	}
+	uclass_foreach_dev(dev, uc) {
+		ret = device_probe(dev);
 		if (ret)
-			printf("%s - probe failed: %d\n", m->name, ret);
+			printf("%s - probe failed: %d\n", dev->name, ret);
 	}
 
 	return 0;