Hi Guoqing,
Sorry for the delay, I missed it.

On Sat, 12 Feb 2022 09:47:38 +0800
Guoqing Jiang [off-list ref] wrote:

On 1/27/22 11:39 PM, Mariusz Tkaczyk wrote:

quoted

Raid456 module had allowed to achieve failed state. It was fixed by
fb73b357fb9 ("raid5: block failing device if raid will be failed").
This fix introduces a bug, now if raid5 fails during IO, it may
result with a hung task without completion. Faulty flag on the
device is necessary to process all requests and is checked many
times, mainly in analyze_stripe().
Allow to set faulty on drive again and set MD_BROKEN if raid is
failed.

As a result, this level is allowed to achieve failed state again,
but communication with userspace (via -EBUSY status) will be
preserved.

This restores possibility to fail array via #mdadm --set-faulty
command and will be fixed by additional verification on mdadm side.
 

Again, you better to send mdadm change along with the series.

I understand your objections. Unfortunately, I was unable to handle it
yet. I focused on kernel part first because mdadm was in freeze.
In mdadm, I need to block manual removal which should be simple, there
is already enough() function defined. It is on top of my TODO lst.
  

quoted

Reproduction steps:
  mdadm -CR imsm -e imsm -n 3 /dev/nvme[0-2]n1
  mdadm -CR r5 -e imsm -l5 -n3 /dev/nvme[0-2]n1 --assume-clean
  mkfs.xfs /dev/md126 -f
  mount /dev/md126 /mnt/root/

  fio --filename=/mnt/root/file --size=5GB --direct=1 --rw=randrw
--bs=64k --ioengine=libaio --iodepth=64 --runtime=240 --numjobs=4
--time_based --group_reporting --name=throughput-test-job
--eta-newline=1 &

  echo 1 > /sys/block/nvme2n1/device/device/remove
  echo 1 > /sys/block/nvme1n1/device/device/remove

  [ 1475.787779] Call Trace:
  [ 1475.793111] __schedule+0x2a6/0x700
  [ 1475.799460] schedule+0x38/0xa0
  [ 1475.805454] raid5_get_active_stripe+0x469/0x5f0 [raid456]
  [ 1475.813856] ? finish_wait+0x80/0x80
  [ 1475.820332] raid5_make_request+0x180/0xb40 [raid456]
  [ 1475.828281] ? finish_wait+0x80/0x80
  [ 1475.834727] ? finish_wait+0x80/0x80
  [ 1475.841127] ? finish_wait+0x80/0x80
  [ 1475.847480] md_handle_request+0x119/0x190
  [ 1475.854390] md_make_request+0x8a/0x190
  [ 1475.861041] generic_make_request+0xcf/0x310
  [ 1475.868145] submit_bio+0x3c/0x160
  [ 1475.874355] iomap_dio_submit_bio.isra.20+0x51/0x60
  [ 1475.882070] iomap_dio_bio_actor+0x175/0x390
  [ 1475.889149] iomap_apply+0xff/0x310
  [ 1475.895447] ? iomap_dio_bio_actor+0x390/0x390
  [ 1475.902736] ? iomap_dio_bio_actor+0x390/0x390
  [ 1475.909974] iomap_dio_rw+0x2f2/0x490
  [ 1475.916415] ? iomap_dio_bio_actor+0x390/0x390
  [ 1475.923680] ? atime_needs_update+0x77/0xe0
  [ 1475.930674] ? xfs_file_dio_aio_read+0x6b/0xe0 [xfs]
  [ 1475.938455] xfs_file_dio_aio_read+0x6b/0xe0 [xfs]
  [ 1475.946084] xfs_file_read_iter+0xba/0xd0 [xfs]
  [ 1475.953403] aio_read+0xd5/0x180
  [ 1475.959395] ? _cond_resched+0x15/0x30
  [ 1475.965907] io_submit_one+0x20b/0x3c0
  [ 1475.972398] __x64_sys_io_submit+0xa2/0x180
  [ 1475.979335] ? do_io_getevents+0x7c/0xc0
  [ 1475.986009] do_syscall_64+0x5b/0x1a0
  [ 1475.992419] entry_SYSCALL_64_after_hwframe+0x65/0xca
  [ 1476.000255] RIP: 0033:0x7f11fc27978d
  [ 1476.006631] Code: Bad RIP value.
  [ 1476.073251] INFO: task fio:3877 blocked for more than 120
seconds.  

Does it also happen to non imsm array? And did you try to reproduce
it with revert fb73b357fb? I suppose fb73b357fb9 introduced the
regression given it is fixed by this one.

It is reproducible on native and IMSM. Yes, we can fix it by revert.

quoted

   * In both cases, &MD_BROKEN will be set in &mddev->flags.
   */
  static void raid1_error(struct mddev *mddev, struct md_rdev *rdev)

diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 1240a5c16af8..bee953c8007f 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c

@@ -686,17 +686,21 @@ int raid5_calc_degraded(struct r5conf *conf)
  	return degraded;
  }
  
-static int has_failed(struct r5conf *conf)
+static bool has_failed(struct r5conf *conf)
  {
-	int degraded;
+	int degraded = conf->mddev->degraded;
  
-	if (conf->mddev->reshape_position == MaxSector)
-		return conf->mddev->degraded > conf->max_degraded;
+	if (test_bit(MD_BROKEN, &conf->mddev->flags))
+		return true;  

If one member disk was set Faulty which caused BROKEN was set, is it
possible to re-add the same member disk again?

Is possible to re-add drive to failed raid5 array now? From my
understanding of raid5_add_disk it is not possible.

[root@vm ~]# echo faulty > /sys/block/md0/md/dev-loop1/state
[root@vm ~]# cat /proc/mdstat
Personalities : [raid6] [raid5] [raid4]
md0 : active raid5 loop2[2] loop1[0](F)
       1046528 blocks super 1.2 level 5, 512k chunk, algorithm 2
[2/1] [_U] bitmap: 0/1 pages [0KB], 65536KB chunk

unused devices: <none>
[root@vm ~]# echo re-add > /sys/block/md0/md/dev-loop1/state
[root@vm ~]# cat /proc/mdstat
Personalities : [raid6] [raid5] [raid4]
md0 : active raid5 loop2[2] loop1[0]
       1046528 blocks super 1.2 level 5, 512k chunk, algorithm 2
[2/2] [UU] bitmap: 0/1 pages [0KB], 65536KB chunk

unused devices: <none>

And have you run mdadm test against the series?

I run imsm test suite and our internal IMSM scope. I will take the
challenge and will verify with native. Thanks for suggestion.

quoted

-	degraded = raid5_calc_degraded(conf);
-	if (degraded > conf->max_degraded)
-		return 1;
-	return 0;
+	if (conf->mddev->reshape_position != MaxSector)
+		degraded = raid5_calc_degraded(conf);
+
+	if (degraded > conf->max_degraded) {
+		set_bit(MD_BROKEN, &conf->mddev->flags);  

Why not set BROKEN flags in err handler to align with other levels? Or
do it in md_error only.

https://lore.kernel.org/linux-raid/3da9324e-01e7-2a07-4bcd-14245db56693@linux.dev/ (local)

You suggested that.
Other levels doesn't have dedicates has_failed() routines. For raid5 it
is reasonable to set it in has_failed().

I can't do that in md_error because I don't have such information in
all cases. !test_bit("Faulty", rdev->flags) result varies.

quoted

+		return true;
+	}
+	return false;
  }
  
  struct stripe_head *

@@ -2877,34 +2881,29 @@ static void raid5_error(struct mddev

*mddev, struct md_rdev *rdev) unsigned long flags;
  	pr_debug("raid456: error called\n");
  
+	pr_crit("md/raid:%s: Disk failure on %s, disabling
device.\n",
+		mdname(mddev), bdevname(rdev->bdev, b));
+
  	spin_lock_irqsave(&conf->device_lock, flags);
+	set_bit(Faulty, &rdev->flags);
+	clear_bit(In_sync, &rdev->flags);
+	mddev->degraded = raid5_calc_degraded(conf);
  
-	if (test_bit(In_sync, &rdev->flags) &&
-	    mddev->degraded == conf->max_degraded) {
-		/*
-		 * Don't allow to achieve failed state
-		 * Don't try to recover this device
-		 */
+	if (has_failed(conf)) {
  		conf->recovery_disabled =
mddev->recovery_disabled;
-		spin_unlock_irqrestore(&conf->device_lock, flags);
-		return;  

Ok, I think commit fb73b357fb985cc652a72a41541d25915c7f9635 is
effectively reverted by this hunk. So I would prefer to separate the
revert part from this patch, just my 0.02$.

Song,
do you want revert?

Thanks,
Mariusz