mdadm/0017-Retry-HOT_REMOVE_DISK-a-few-times.patch

From 2dd271fe7041c2f7036939cf6917c0578b92fefe Mon Sep 17 00:00:00 2001
From: NeilBrown <neilb@suse.com>
Date: Mon, 27 Mar 2017 12:50:16 +1100
Subject: [PATCH] Retry HOT_REMOVE_DISK a few times.
References: bsc#808647

HOT_REMOVE_DISK can fail with EBUSY if there are outstanding
IO request that have not completed yet.  It can sometimes
be helpful to wait a little while for these to complete.

We already do this in impose_level() when reshaping a device,
but not in Manage.c in response to an explicit --remove request.

So create hot_remove_disk() to central this code, and call it
where-ever it makes sense to wait for a HOT_REMOVE_DISK to succeed.

Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Jes Sorensen <Jes.Sorensen@gmail.com>
---
 Grow.c   |  9 +--------
 Manage.c |  4 ++--
 mdadm.h  |  1 +
 util.c   | 18 ++++++++++++++++++
 4 files changed, 22 insertions(+), 10 deletions(-)

diff --git a/Grow.c b/Grow.c
index 455c5f9..218a706 100755
--- a/Grow.c
+++ b/Grow.c
@@ -2736,7 +2736,6 @@ static int impose_level(int fd, int level, char *devname, int verbose)
 		for (d = 0, found = 0;
 		     d < MAX_DISKS && found < array.nr_disks;
 		     d++) {
-			int cnt;
 			mdu_disk_info_t disk;
 			disk.number = d;
 			if (ioctl(fd, GET_DISK_INFO, &disk) < 0)
@@ -2750,13 +2749,7 @@ static int impose_level(int fd, int level, char *devname, int verbose)
 				continue;
 			ioctl(fd, SET_DISK_FAULTY,
 			      makedev(disk.major, disk.minor));
-			cnt = 5;
-			while (ioctl(fd, HOT_REMOVE_DISK,
-				     makedev(disk.major, disk.minor)) < 0
-			       && errno == EBUSY
-			       && cnt--) {
-				usleep(10000);
-			}
+			hot_remove_disk(fd, makedev(disk.major, disk.minor));
 		}
 	}
 	c = map_num(pers, level);
diff --git a/Manage.c b/Manage.c
index 5c3d2b9..9139f96 100644
--- a/Manage.c
+++ b/Manage.c
@@ -1183,7 +1183,7 @@ int Manage_remove(struct supertype *tst, int fd, struct mddev_dev *dv,
 		else
 			err = 0;
 	} else {
-		err = ioctl(fd, HOT_REMOVE_DISK, rdev);
+		err = hot_remove_disk(fd, rdev);
 		if (err && errno == ENODEV) {
 			/* Old kernels rejected this if no personality
 			 * is registered */
@@ -1607,7 +1607,7 @@ int Manage_subdevs(char *devname, int fd,
 
 			if (dv->disposition == 'F')
 				/* Need to remove first */
-				ioctl(fd, HOT_REMOVE_DISK, rdev);
+				hot_remove_disk(fd, rdev);
 			/* Make sure it isn't in use (in 2.6 or later) */
 			tfd = dev_open(dv->devname, O_RDONLY|O_EXCL);
 			if (tfd >= 0) {
diff --git a/mdadm.h b/mdadm.h
index 91fd9eb..5bcfb86 100644
--- a/mdadm.h
+++ b/mdadm.h
@@ -1476,6 +1476,7 @@ extern int add_disk(int mdfd, struct supertype *st,
 		    struct mdinfo *sra, struct mdinfo *info);
 extern int remove_disk(int mdfd, struct supertype *st,
 		       struct mdinfo *sra, struct mdinfo *info);
+extern int hot_remove_disk(int mdfd, unsigned long dev);
 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
 unsigned long long min_recovery_start(struct mdinfo *array);
 
diff --git a/util.c b/util.c
index 32bd909..d09a7e2 100644
--- a/util.c
+++ b/util.c
@@ -1795,6 +1795,24 @@ int remove_disk(int mdfd, struct supertype *st,
 	return rv;
 }
 
+int hot_remove_disk(int mdfd, unsigned long dev)
+{
+	int cnt = 5;
+	int ret;
+
+	/* HOT_REMOVE_DISK can fail with EBUSY if there are
+	 * outstanding IO requests to the device.
+	 * In this case, it can be helpful to wait a little while,
+	 * up to half a second, for that IO to flush.
+	 */
+	while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 &&
+	       errno == EBUSY &&
+	       cnt-- > 0)
+		usleep(10000);
+
+	return ret;
+}
+
 int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
 {
 	/* Initialise kernel's knowledge of array.
-- 
2.10.2