On Fri, 3 Dec 2010 07:45:14 +0000 "Kwolek, Adam" [off-list ref]
wrote:

I think you applied not a whole patch (no 't' definition and initialization), you missed:

+		struct timeval t;
+
+		t.tv_sec = 1;
+		t.tv_usec = 0;

so mdadm is not compiling

Thanks.

I meant to remove the 'struct timeval t', but I forgot to remove the 't' from
select.
And somehow I forgot to run 'make' after that patch.

Fixed now.
Thanks,
NeilBrown

BR
Adam

quoted

-----Original Message-----
From: Neil Brown [mailto:neilb@suse.de]
Sent: Friday, December 03, 2010 5:17 AM
To: Kwolek, Adam
Cc: linux-raid@vger.kernel.org; Williams, Dan J; Ciechanowski, Ed
Subject: Re: [PATCH 10/10] FIX: wait_backup() sometimes hungs

On Thu, 02 Dec 2010 09:19:58 +0100 Adam Kwolek [off-list ref]
wrote:

quoted

Sometimes wait_backup() omits transition from reshape to iddle state

and mdadm seams to be hung.

quoted

Add 1 sec. timeout wor waiting on select. This allows for wait_backup

exit when reshape is ended.

quoted

Signed-off-by: Adam Kwolek <redacted>
---

 Grow.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/Grow.c b/Grow.c
index 24c5c39..e16b1ad 100644
--- a/Grow.c
+++ b/Grow.c

@@ -2074,10 +2074,14 @@ static int wait_backup(struct mdinfo *sra,
 		sysfs_set_str(sra, NULL, "sync_action", "reshape");
 	do {
 		char action[20];
+		struct timeval t;
+
+		t.tv_sec = 1;
+		t.tv_usec = 0;
 		fd_set rfds;
 		FD_ZERO(&rfds);
 		FD_SET(fd, &rfds);
-		select(fd+1, NULL, NULL, &rfds, NULL);
+		select(fd+1, NULL, NULL, &rfds, &t);
 		if (sysfs_fd_get_ll(fd, &completed) < 0) {
 			close(fd);
 			return -1;

Thanks.  However I don't think the 1 second timeout is necessary.  This
is
really the same problem as the previous one.  We just need to read
'completed' before the first 'select'.  Like this.

Thanks,
NeilBrown

commit 97bef35459306dfd291f40bc5221ad20ab9c21ba
Author: Adam Kwolek [off-list ref]
Date:   Fri Dec 3 15:15:51 2010 +1100

    FIX: wait_backup() sometimes hungs

    Sometimes wait_backup() omits transition from reshape to idle state
    and mdadm seams to be hung.  So check the 'complete' count
    *before* waiting rather than only after.

    Signed-off-by: Adam Kwolek [off-list ref]
    Signed-off-by: NeilBrown [off-list ref]

diff --git a/Grow.c b/Grow.c
index 3322cf7..99807b4 100644
--- a/Grow.c
+++ b/Grow.c

@@ -2058,12 +2058,17 @@ static int wait_backup(struct mdinfo *sra,
 	sysfs_set_num(sra, NULL, "sync_max", offset + blocks + blocks2);
 	if (offset == 0)
 		sysfs_set_str(sra, NULL, "sync_action", "reshape");
-	do {
+
+	if (sysfs_fd_get_ll(fd, &completed) < 0) {
+		close(fd);
+		return -1;
+	}
+	while (completed < offset + blocks) {
 		char action[20];
 		fd_set rfds;
 		FD_ZERO(&rfds);
 		FD_SET(fd, &rfds);
-		select(fd+1, NULL, NULL, &rfds, NULL);
+		select(fd+1, NULL, NULL, &rfds, &t);
 		if (sysfs_fd_get_ll(fd, &completed) < 0) {
 			close(fd);
 			return -1;

@@ -2072,7 +2077,7 @@ static int wait_backup(struct mdinfo *sra,
 				  action, 20) > 0 &&
 		    strncmp(action, "reshape", 7) != 0)
 			break;
-	} while (completed < offset + blocks);
+	}
 	close(fd);

 	if (part) {