Changes from V2:
- Fix to prevent the array from being marked broken for all
Failfast IOs, not just metadata.
- Reflecting the review, update raid{1,10}_error to clear
FailfastIOFailure so that devices are properly marked Faulty.
Changes from V1:
- Avoid setting MD_BROKEN instead of clearing it
- Add pr_crit() when setting MD_BROKEN
- Fix the message may shown after all rdevs failure:
"Operation continuing on 0 devices"
v2: https://lore.kernel.org/linux-raid/20250817172710.4892-1-k@xxxxxxx/
v1: https://lore.kernel.org/linux-raid/20250812090119.153697-1-k@xxxxxxx/
A failfast bio, for example in the case of nvme-tcp, bio will fail
immediately if the connection to the target is briefly lost and
the device enters a reconnecting state - even though it would
recover given few seconds. This behavior is by design in failfast.
However, md treats Failfast IO failures as fatal,
potentially marking the array as MD_BROKEN when a connection is lost.
For example, if an initiator - that is, a machine loading the md
module - loses all connections briefly, the array is marked
as MD_BROKEN, preventing subsequent writes.
This is the issue I am currently facing, and which this patch aims to fix.
The 1st patch changes the behavior on MD_FAILFAST IO failures on
the last rdev. The 2nd and 3rd patches modify the pr_crit messages.
Kenta Akagi (3):
md/raid1,raid10: Do not set MD_BROKEN on failfast io failure
md/raid1,raid10: Add error message when setting MD_BROKEN
md/raid1,raid10: Fix: Operation continuing on 0 devices.
drivers/md/md.c | 14 +++++++++-----
drivers/md/md.h | 13 +++++++------
drivers/md/raid1.c | 32 ++++++++++++++++++++++++++------
drivers/md/raid10.c | 35 ++++++++++++++++++++++++++++-------
4 files changed, 70 insertions(+), 24 deletions(-)
--
2.50.1
