Thread (4 messages) 4 messages, 2 authors, 2011-01-17

RE: [PATCH 2/2] Assemble: allow to assemble spares on their own

flat view

From: Czarnowska, Anna <hidden>
Date: 2011-01-09 22:29:51

-----Original Message-----
From: linux-raid-owner@vger.kernel.org [mailto:linux-raid-
owner@vger.kernel.org] On Behalf Of NeilBrown
Sent: Wednesday, January 05, 2011 3:59 AM
To: Czarnowska, Anna
Cc: linux-raid@vger.kernel.org; Williams, Dan J; Ciechanowski, Ed;
Hawrylewicz Czarnowski, Przemyslaw; Labun, Marcin; Neubauer, Wojciech
Subject: Re: [PATCH 2/2] Assemble: allow to assemble spares on their
own

On Tue, 4 Jan 2011 14:06:30 +0000 "Czarnowska, Anna"
[off-list ref] wrote:
quoted
quoted
From eb737197e6a3a43f6724003900a87410acec9495 Mon Sep 17 00:00:00
2001
quoted
From: Anna Czarnowska <redacted>
Date: Tue, 4 Jan 2011 12:18:18 +0100
Subject: [PATCH 2/2] Assemble: allow to assemble spares on their own
Cc: linux-raid@vger.kernel.org, Williams, Dan J
[off-list ref], Ciechanowski, Ed
[off-list ref]
quoted
If we find spares but no members of given array
we create container with just spares.

This allows auto assemble to pick up all lose imsm spares when there
is no config file.
When there is a valid config file and any array is assembled from it
we don't try auto assembly so we will not assemble spares that don't
match any array.
To remedy this we must add
ARRAY metadata=imsm UUID=00000000:00000000:00000000:00000000
to config file.
This container will include all remaining spares.
Thanks.
I have made some modifications resulting in the following patch.

The main problem was that your code explicitly mentioned "imsm" in
generic
code.  That is always the wrong thing to do.  One day we might have a
different metadata type that has the same sort of floating spare, in
which
case the imsm spares need to be handled separately to the new type of
spares.

So:
 I move the place where ->used is set to '3' so that st gets set.
So if we see and IMSM device, we will create an IMSM array at that
point,
though it might end up being a container of spares.

I removed the 'else' so that the original loop searching for 'used ==
3'
always runs. If we just set up an array of spares, then it will find
nothing,
but that is no loss.  This avoids the need to indent it further.

As mentioned earlier, I added a test for 'auto_assem'.
Adding such test means that we can only assemble spares on their own when there is no config file.
When there is a config file with at least one ARRAY line we do not try auto assembly.
This also means that adding special ARRAY line for spares will not work - in this case
we have no auto_assem (because uuid is set) so we try the second loop where all spares get rejected because domains is null.
Should `mdadm -As` only assemble what it finds in config?
Maybe we should always try auto assembly? 
What is the original reasoning behind not trying auto assembly when there is at least one array in config?
Regards
Anna
quoted hunk ↗ jump to hunk
Thanks,
NeilBrown

commit ed7fc6b4d91a5342886ba6fe6d175f69f0875d3b
Author: Anna Czarnowska [off-list ref]
Date:   Wed Jan 5 13:54:18 2011 +1100

    Assemble: allow to assemble spares on their own

    If we find spares but no members of given array
    we create container with just spares.

    This allows auto assemble to pick up all lose imsm spares when
there
    is no config file.
    When there is a valid config file and any array is assembled from
it
    we don't try auto assembly so we will not assemble spares that
don't
    match any array.
    To remedy this we must add
    ARRAY metadata=imsm UUID=00000000:00000000:00000000:00000000
    to config file.
    This container will include all remaining spares.

    Signed-off-by: Anna Czarnowska [off-list ref]
    Signed-off-by: NeilBrown [off-list ref]
diff --git a/Assemble.c b/Assemble.c
index 7ef9dc3..532335e 100644
--- a/Assemble.c
+++ b/Assemble.c
@@ -489,17 +489,23 @@ int Assemble(struct supertype *st, char *mddev,
 					   homehost, update,
 					   report_missmatch ? devname : NULL))
 				goto loop;
-
-			if (!memcmp(content->uuid, uuid_zero,
sizeof(int[4]))) {
-				/* this is imsm_spare - do not set st */
-				tmpdev->used = 3;
-				goto loop;
-			}

 			if (st == NULL)
 				st = dup_super(tst);
 			if (st->minor_version == -1)
 				st->minor_version = tst->minor_version;
+
+			if (memcmp(content->uuid, uuid_zero,
+				   sizeof(int[4])) == 0) {
+				/* this is a floating spare.  It cannot define
+				 * an array unless there are no more arrays of
+				 * this type to be found.  It can be included
+				 * in an array of this type though.
+				 */
+				tmpdev->used = 3;
+				goto loop;
+			}
+
 			if (st->ss != tst->ss ||
 			    st->minor_version != tst->minor_version ||
 			    st->ss->compare_super(st, tst) != 0) {
@@ -557,6 +563,25 @@ int Assemble(struct supertype *st, char *mddev,
 			tst->ss->free_super(tst);
 	}

+	/* Check if we found some imsm spares but no members */
+	if (auto_assem && (!st || !st->sb))
+		for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
+			if (tmpdev->used != 3)
+				continue;
+			tmpdev->used = 1;
+			content = &info;
+
+			if (!st->sb) {
+				/* we need sb from one of the spares */
+				int dfd = dev_open(tmpdev->devname, O_RDONLY);
+				if (dfd < 0 ||
+				    st->ss->load_super(st, dfd, NULL))
+					tmpdev->used = 2;
+				if (dfd > 0)
+					close(dfd);
+			}
+		}
+
 	/* Now reject spares that don't match domains of identified
members */
 	for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
 		struct stat stb;
@@ -892,8 +917,10 @@ int Assemble(struct supertype *st, char *mddev,
 		if (content->array.level != LEVEL_MULTIPATH)
 			if (!(devices[j].i.disk.state & (1<<MD_DISK_ACTIVE)))
{
 				if (!(devices[j].i.disk.state
-				      & (1<<MD_DISK_FAULTY)))
+				      & (1<<MD_DISK_FAULTY))) {
+					devices[j].uptodate = 1;
 					sparecnt++;
+				}
 				continue;
 			}
 		/* If this devices thinks that 'most_recent' has failed,
then
--
To unsubscribe from this list: send the line "unsubscribe linux-raid"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help