Re: [PATCH -next v3 4/7] md: factor out a helper rdev_removeable() from remove_and_add_spares()

From: Yu Kuai
Date: Tue Aug 22 2023 - 22:45:24 EST


Hi,

在 2023/08/22 18:19, Xiao Ni 写道:
On Sun, Aug 20, 2023 at 5:13 PM Yu Kuai <yukuai1@xxxxxxxxxxxxxxx> wrote:

From: Yu Kuai <yukuai3@xxxxxxxxxx>

There are no functional changes, just to make the code simpler and
prepare to delay remove_and_add_spares() to md_start_sync().

Signed-off-by: Yu Kuai <yukuai3@xxxxxxxxxx>
---
drivers/md/md.c | 33 +++++++++++++++++++--------------
1 file changed, 19 insertions(+), 14 deletions(-)

diff --git a/drivers/md/md.c b/drivers/md/md.c
index 561cac13ff96..ceace5ffadd6 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -9153,6 +9153,22 @@ void md_do_sync(struct md_thread *thread)
}
EXPORT_SYMBOL_GPL(md_do_sync);

+static bool rdev_removeable(struct md_rdev *rdev)
+{
+ if (rdev->raid_disk < 0 || test_bit(Blocked, &rdev->flags) ||
+ atomic_read(&rdev->nr_pending))
+ return false;
+
+ if (test_bit(RemoveSynchronized, &rdev->flags))
+ return true;
+
+ if (test_bit(In_sync, &rdev->flags) ||
+ test_bit(Journal, &rdev->flags))
+ return false;
+
+ return true;
+}
+
static int remove_and_add_spares(struct mddev *mddev,
struct md_rdev *this)
{
@@ -9166,11 +9182,7 @@ static int remove_and_add_spares(struct mddev *mddev,
return 0;

rdev_for_each(rdev, mddev) {
- if ((this == NULL || rdev == this) &&
- rdev->raid_disk >= 0 &&
- !test_bit(Blocked, &rdev->flags) &&
- test_bit(Faulty, &rdev->flags) &&
- atomic_read(&rdev->nr_pending)==0) {
+ if ((this == NULL || rdev == this) && rdev_removeable(rdev)) {

There is a small change with the original method. Before this patch,
it checks the Faulty flag when setting RemoveSynchronized and it
checks RemoveSynchronized and "!In_sync && !Journal". I'm not sure if
it's right or not.

Yes, there is a small change. After a second thought, I think it's OK
to leave the code to set RemoveSynchronized where it is for now, because
it'll be removed later. I don't need to bother factor out a common code
to set RemoveSynchronized and call hot_remove_disk().

By the way, once refactor of mddev_suspend() is done, then access to
rdev from fastpath will be replaced from:

rcu_read_lock()
...
rcu_read_unlock()

to:

md_array_enter()
// grab 'active_io', 'active_io' will probably be renamed
...
md_array_exit()

That's why I said RemoveSynchronized will be removed.

Thanks,
Kuai


/* Faulty non-Blocked devices with nr_pending == 0
* never get nr_pending incremented,
* never get Faulty cleared, and never get Blocked set.
@@ -9185,19 +9197,12 @@ static int remove_and_add_spares(struct mddev *mddev,
synchronize_rcu();
rdev_for_each(rdev, mddev) {
if ((this == NULL || rdev == this) &&
- rdev->raid_disk >= 0 &&
- !test_bit(Blocked, &rdev->flags) &&
- ((test_bit(RemoveSynchronized, &rdev->flags) ||
- (!test_bit(In_sync, &rdev->flags) &&
- !test_bit(Journal, &rdev->flags))) &&
- atomic_read(&rdev->nr_pending)==0)) {
- if (mddev->pers->hot_remove_disk(
- mddev, rdev) == 0) {
+ rdev_removeable(rdev) &&
+ mddev->pers->hot_remove_disk(mddev, rdev) == 0) {
sysfs_unlink_rdev(mddev, rdev);
rdev->saved_raid_disk = rdev->raid_disk;
rdev->raid_disk = -1;
removed++;
- }
}
if (remove_some && test_bit(RemoveSynchronized, &rdev->flags))
clear_bit(RemoveSynchronized, &rdev->flags);
--
2.39.2


.