mdadm/0104-imsm-update-num_data_stripes-according-to-dev_size.patch

From 895ffd992954069e4ea67efb8a85bb0fd72c3707 Mon Sep 17 00:00:00 2001
From: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Date: Tue, 24 Nov 2020 14:15:15 +0100
Subject: [PATCH 10/17] imsm: update num_data_stripes according to dev_size
Git-commit: 895ffd992954069e4ea67efb8a85bb0fd72c3707
References: jsc#SLE-13700

If array was created in UEFI there is possibility that
member size is not rounded to 1MB. After any size reconfiguration
it will be rounded down to 1MB per each member but the old
component size will remain in metadata.
During reshape old array size is calculated from component size because
dev_size is not a part of map and is bumped to new value quickly.
It may result in size mismatch if array is assembled during reshape.

If difference in calculated size and dev_size is observed try to fix it.
num_data_stripes value can be safety updated to smaller value if array
doesn't occuppy whole reserved component space.

Signed-off-by: Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>
Signed-off-by: Coly Li <colyli@suse.de>

---
 super-intel.c | 84 +++++++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 78 insertions(+), 6 deletions(-)

diff --git a/super-intel.c b/super-intel.c
index 3a73d2b..9562064 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -3453,7 +3453,6 @@ static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info,
 			__u64 blocks_per_unit = blocks_per_migr_unit(super,
 								     dev);
 			__u64 units = current_migr_unit(migr_rec);
-			unsigned long long array_blocks;
 			int used_disks;
 
 			if (__le32_to_cpu(migr_rec->ascending_migr) &&
@@ -3472,12 +3471,8 @@ static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info,
 
 			used_disks = imsm_num_data_members(prev_map);
 			if (used_disks > 0) {
-				array_blocks = per_dev_array_size(map) *
+				info->custom_array_size = per_dev_array_size(map) *
 					used_disks;
-				info->custom_array_size =
-					round_size_to_mb(array_blocks,
-							 used_disks);
-
 			}
 		}
 		case MIGR_VERIFY:
@@ -11682,6 +11677,68 @@ int imsm_takeover(struct supertype *st, struct geo_params *geo)
 	return 0;
 }
 
+/* Flush size update if size calculated by num_data_stripes is higher than
+ * imsm_dev_size to eliminate differences during reshape.
+ * Mdmon will recalculate them correctly.
+ * If subarray index is not set then check whole container.
+ * Returns:
+ *	0 - no error occurred
+ *	1 - error detected
+ */
+static int imsm_fix_size_mismatch(struct supertype *st, int subarray_index)
+{
+	struct intel_super *super = st->sb;
+	int tmp = super->current_vol;
+	int ret_val = 1;
+	int i;
+
+	for (i = 0; i < super->anchor->num_raid_devs; i++) {
+		if (subarray_index >= 0 && i != subarray_index)
+			continue;
+		super->current_vol = i;
+		struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
+		struct imsm_map *map = get_imsm_map(dev, MAP_0);
+		unsigned int disc_count = imsm_num_data_members(map);
+		struct geo_params geo;
+		struct imsm_update_size_change *update;
+		unsigned long long calc_size = per_dev_array_size(map) * disc_count;
+		unsigned long long d_size = imsm_dev_size(dev);
+		int u_size;
+
+		if (calc_size == d_size || dev->vol.migr_type == MIGR_GEN_MIGR)
+			continue;
+
+		/* There is a difference, verify that imsm_dev_size is
+		 * rounded correctly and push update.
+		 */
+		if (d_size != round_size_to_mb(d_size, disc_count)) {
+			dprintf("imsm: Size of volume %d is not rounded correctly\n",
+				 i);
+			goto exit;
+		}
+		memset(&geo, 0, sizeof(struct geo_params));
+		geo.size = d_size;
+		u_size = imsm_create_metadata_update_for_size_change(st, &geo,
+								     &update);
+		if (u_size < 1) {
+			dprintf("imsm: Cannot prepare size change update\n");
+			goto exit;
+		}
+		imsm_update_metadata_locally(st, update, u_size);
+		if (st->update_tail) {
+			append_metadata_update(st, update, u_size);
+			flush_metadata_updates(st);
+			st->update_tail = &st->updates;
+		} else {
+			imsm_sync_metadata(st);
+		}
+	}
+	ret_val = 0;
+exit:
+	super->current_vol = tmp;
+	return ret_val;
+}
+
 static int imsm_reshape_super(struct supertype *st, unsigned long long size,
 			      int level,
 			      int layout, int chunksize, int raid_disks,
@@ -11718,6 +11775,11 @@ static int imsm_reshape_super(struct supertype *st, unsigned long long size,
 			struct imsm_update_reshape *u = NULL;
 			int len;
 
+			if (imsm_fix_size_mismatch(st, -1)) {
+				dprintf("imsm: Cannot fix size mismatch\n");
+				goto exit_imsm_reshape_super;
+			}
+
 			len = imsm_create_metadata_update_for_reshape(
 				st, &geo, old_raid_disks, &u);
 
@@ -12020,6 +12082,7 @@ static int imsm_manage_reshape(
 	unsigned long long start_buf_shift; /* [bytes] */
 	int degraded = 0;
 	int source_layout = 0;
+	int subarray_index = -1;
 
 	if (!sra)
 		return ret_val;
@@ -12033,6 +12096,7 @@ static int imsm_manage_reshape(
 		    dv->dev->vol.migr_state == 1) {
 			dev = dv->dev;
 			migr_vol_qan++;
+			subarray_index = dv->index;
 		}
 	}
 	/* Only one volume can migrate at the same time */
@@ -12217,6 +12281,14 @@ static int imsm_manage_reshape(
 
 	/* return '1' if done */
 	ret_val = 1;
+
+	/* After the reshape eliminate size mismatch in metadata.
+	 * Don't update md/component_size here, volume hasn't
+	 * to take whole space. It is allowed by kernel.
+	 * md/component_size will be set propoperly after next assembly.
+	 */
+	imsm_fix_size_mismatch(st, subarray_index);
+
 abort:
 	free(buf);
 	/* See Grow.c: abort_reshape() for further explanation */
-- 
2.26.2