[v1,0/3] WriteBooster Feature Support

Message

Asutosh Das (asd) April 8, 2020, 9:48 p.m. UTC

RFC -> v1:
- Added platform capability for WriteBooster 

RFC-:
v1 -> v2:
- Addressed comments on v1

- Supports shared buffer mode only

- Didn't use exception event as suggested.
  The reason being while testing I saw that the WriteBooster
  available buffer remains at 0x1 for a longer time if flush is
  enabled all the time as compared to an event-based enablement.
  This essentially means that writes go to the WriteBooster buffer
  more. Spec says that the if flush is enabled, the device would
  flush when it sees the command queue empty. So I guess that'd trigger
  flush more than an event based approach.
  Anyway the Vcc would be turned-off during system suspend, so flush
  would stop anyway.
  In this patchset, I never turn-off flush.
  Hence the RFC.

Asutosh Das (3):
  scsi: ufs: add write booster feature support
  ufs-qcom: scsi: configure write booster type
  ufs: sysfs: add sysfs entries for write booster

 drivers/scsi/ufs/ufs-qcom.c  |   8 ++
 drivers/scsi/ufs/ufs-sysfs.c |  39 ++++++-
 drivers/scsi/ufs/ufs.h       |  37 ++++++-
 drivers/scsi/ufs/ufshcd.c    | 241 ++++++++++++++++++++++++++++++++++++++++++-
 drivers/scsi/ufs/ufshcd.h    |  43 ++++++++
 5 files changed, 362 insertions(+), 6 deletions(-)

Comments

Avri Altman April 23, 2020, 8:30 a.m. UTC | #1

Looks good.
Reviewed-by: Avri Altman <[e-mail]>

Thanks,
Avri

> v2 -> v3:
> - Addressed Comments on refactoring
> - Corrected the commit message
> 
> v1 -> v2:
> - Refactor WriteBooster initialization, introduce ufshcd_wb_probe()
> - Refactor ufshcd_wb_keep_vcc_on() and introduce a new function
>   ufshcd_wb_presrv_usrspc_keep_vcc_on()
> - Get the WriteBooster configuration by reading
>   bWriteBoosterBufferPreserveUserSpaceEn
> 
> RFC -> v1:
> - Added platform capability for WriteBooster
> 
> RFC-:
> v1 -> v2:
> - Addressed comments on v1
> 
> - Supports shared buffer mode only
> 
> - Didn't use exception event as suggested.
>   The reason being while testing I saw that the WriteBooster
>   available buffer remains at 0x1 for a longer time if flush is
>   enabled all the time as compared to an event-based enablement.
>   This essentially means that writes go to the WriteBooster buffer
>   more. Spec says that the if flush is enabled, the device would
>   flush when it sees the command queue empty. So I guess that'd trigger
>   flush more than an event based approach.
>   Anyway the Vcc would be turned-off during system suspend, so flush
>   would stop anyway.
>   In this patchset, I never turn-off flush.
>   Hence the RFC.
> 
> Asutosh Das (3):
>   scsi: ufs: add write booster feature support
>   ufs-qcom: scsi: configure write booster type
>   ufs: sysfs: add sysfs entries for write booster
> 
>  drivers/scsi/ufs/ufs-qcom.c  |   8 ++
>  drivers/scsi/ufs/ufs-sysfs.c |  39 ++++++-
>  drivers/scsi/ufs/ufs.h       |  37 ++++++-
>  drivers/scsi/ufs/ufshcd.c    | 241
> ++++++++++++++++++++++++++++++++++++++++++-
>  drivers/scsi/ufs/ufshcd.h    |  43 ++++++++
>  5 files changed, 362 insertions(+), 6 deletions(-)
> 
> --
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a
> Linux Foundation Collaborative Project.

Martin K. Petersen April 27, 2020, 10:49 p.m. UTC | #2

On Wed, 8 Apr 2020 14:48:38 -0700, Asutosh Das wrote:

Applied to 5.8/scsi-queue, thanks!

[1/3] scsi: ufs: Add write booster feature support
      https://git.kernel.org/mkp/scsi/c/3d17b9b5ab11
[2/3] scsi: ufs: sysfs: Add sysfs entries for write booster
      https://git.kernel.org/mkp/scsi/c/c14e7adf3a6a
[3/3] scsi: ufs-qcom: Configure write booster type
      https://git.kernel.org/mkp/scsi/c/04ee8a01abf8