mbox series

[0/7] Fix UBI Block wrt. highmem

Message ID 20230810160019.16977-1-richard@nod.at (mailing list archive)
Headers show
Series Fix UBI Block wrt. highmem | expand

Message

Richard Weinberger Aug. 10, 2023, 4 p.m. UTC
Patch 1 changes UBIblock to use a copy of scsi_kmap_atomic_sg()
for sg list processing. This patch is meant for backporting to stable.
It makes use of kmap_atomic() and a bounce buffer because MTD/UBI IO
can sleep.

Patches 2 to 7 move scsi_kmap_atomic_sg() into lib/scatterlist.c,
convert it to kmap_local(), convert all users to it and remove the
bounce buffer from UBIblock again.

Richard Weinberger (7):
  ubi: block: Refactor sg list processing for highmem
  scatterlist: Add kmap helpers
  scsi: dc395x: Switch to kmap_sg
  scsi: esp_scsi: Switch to kmap_sg
  scsi: fdomain: Switch to kmap_sg
  ubi: block: Switch to kmap_sg
  scsi: core: Remove scsi_kmap_atomic_sg()

 drivers/mtd/ubi/block.c     | 11 +++----
 drivers/mtd/ubi/eba.c       | 50 +++++++++++++------------------
 drivers/scsi/dc395x.c       | 12 ++++----
 drivers/scsi/esp_scsi.c     |  4 +--
 drivers/scsi/fdomain.c      | 10 +++----
 drivers/scsi/scsi_lib.c     | 60 -------------------------------------
 include/linux/mtd/ubi.h     | 12 ++++----
 include/linux/scatterlist.h |  3 ++
 include/scsi/scsi_cmnd.h    |  4 ---
 lib/scatterlist.c           | 55 ++++++++++++++++++++++++++++++++++
 10 files changed, 100 insertions(+), 121 deletions(-)

Comments

Miquel Raynal Aug. 11, 2023, 5:52 a.m. UTC | #1
Hi Richard,

richard@nod.at wrote on Thu, 10 Aug 2023 18:00:11 +0200:

> Patch 1 changes UBIblock to use a copy of scsi_kmap_atomic_sg()
> for sg list processing. This patch is meant for backporting to stable.
> It makes use of kmap_atomic() and a bounce buffer because MTD/UBI IO
> can sleep.
> 
> Patches 2 to 7 move scsi_kmap_atomic_sg() into lib/scatterlist.c,
> convert it to kmap_local(), convert all users to it and remove the
> bounce buffer from UBIblock again.

Both the idea and the implementation look nice, I don't feel skilled
enough for sending Reviewed-by here, but it seems okay at a first
glance.

Well done Richard :-)

Miquèl