[v9,0/6] Improve libata support for FUA

Message

Damien Le Moal Jan. 10, 2023, 1:27 p.m. UTC

These patches cleanup and improve libata support for ATA devices
supporting the FUA feature.

The first patch modifies the block layer to prevent the use of REQ_FUA
with read requests. This is necessary as the block layer code expect
REQ_FUA to be used with write requests (the flush machinery cannot
enforce access to the media for FUA read commands) and FUA is not
supported with ATA devices when NCQ is not enabled (device queue depth
set to 1).

Patch 2 and 3 are libata cleanup preparatory patches. Patch 4 cleans up
the detection for FUA support. Patch 5 fixes building a taskfile for FUA
write requests. Patch 6 prevents the use of FUA with known bad drives.

Changes from v8:
 - Added Niklas Review tag to patch 1 (forgot to do that when I sent
   v8).

Changes from v7:
 - Given that there are concerns about introducing regressions with
   older devices by enabling FUA by default, patch 7 is dropped for now.
   A more restrictive approach enabling FUA by default only for very
   recent drives will be sent later.

Changes from v6:
 - Modified patch 1 to include checks for REQ_OP_ZONE_APPEND
 - Addressed comments from Niklas (patch 2 -> return false, patch 3 ->
   commit message typo, patch 7 -> more verbose commit message)

Changes from v5:
 - Removed WARN for FUA reads in patch 5.
 - Added reviewed-by tags.

Changes from v4:
 - Changed patch 1 to the one suggested by Christoph.
 - Added Hannes review tag.

Changes from v3:
 - Added patch 1 to prevent any block device user from issuing a
   REQ_FUA read.
 - Changed patch 5 to remove the check for REQ_FUA read and also remove 
   support for ATA_CMD_WRITE_MULTI_FUA_EXT as this command is obsolete
   in recent ACS specifications.

Changes from v2:
 - Added patch 1 and 2 as preparatory patches
 - Added patch 4 to fix FUA writes handling for the non-ncq case. Note
   that it is possible that the drives blacklisted in patch 5 are
   actually OK since the code back in 2012 had the issue with the wrong
   use of LBA 28 commands for FUA writes.

Changes from v1:
 - Removed Maciej's patch 2. Instead, blacklist drives which are known
   to have a buggy FUA support.

Christoph Hellwig (1):
  block: add a sanity check for non-write flush/fua bios

Damien Le Moal (5):
  ata: libata: Introduce ata_ncq_supported()
  ata: libata: Rename and cleanup ata_rwcmd_protocol()
  ata: libata: cleanup fua support detection
  ata: libata: Fix FUA handling in ata_build_rw_tf()
  ata: libata: blacklist FUA support for known buggy drives

 .../admin-guide/kernel-parameters.txt         |  3 +
 block/blk-core.c                              | 14 ++--
 drivers/ata/libata-core.c                     | 69 ++++++++++++++-----
 drivers/ata/libata-scsi.c                     | 30 +-------
 include/linux/libata.h                        | 36 +++++++---
 5 files changed, 92 insertions(+), 60 deletions(-)