Re: [PATCH v3 00/10] lightnvm: next set of improvements for 5.2

["Matias Bjørling" <mb@lightnvm.io>]

On 3/27/19 1:38 PM, Igor Konopko wrote:
> This is another set of fixes and improvements to both pblk and lightnvm
> core.
> 
> First & second patches are the most crutial, since they changes the approach
> to the partial read path, so detailed review is needed especially here.
> 
> Other patches are my other findings related to some bugs or potential
> improvements, mostly related to some corner cases.
> 
> Changes v2 -> v3;
> -dropped some not needed patches
> -dropped patches which were already pulled into for-5.2/core branch
> -commit messages cleanup
> 
> Changes v1 -> v2:
> -dropped some not needed patches
> -review feedback incorporated for some of the patches
> -partial read path changes patch splitted into two patches
> 
> Igor Konopko (10):
>    lightnvm: pblk: IO path reorganization
>    lightnvm: pblk: simplify partial read path
>    lightnvm: pblk: propagate errors when reading meta
>    lightnvm: pblk: recover only written metadata
>    lightnvm: pblk: wait for inflight IOs in recovery
>    lightnvm: pblk: remove internal IO timeout
>    lightnvm: pblk: store multiple copies of smeta
>    lightnvm: pblk: GC error handling
>    lightnvm: pblk: use nvm_rq_to_ppa_list()
>    lightnvm: track inflight target creations
> 
>   drivers/lightnvm/core.c          |  19 ++-
>   drivers/lightnvm/pblk-cache.c    |   8 +-
>   drivers/lightnvm/pblk-core.c     | 179 +++++++++++++++++----
>   drivers/lightnvm/pblk-gc.c       |   5 +-
>   drivers/lightnvm/pblk-init.c     |  37 +++--
>   drivers/lightnvm/pblk-read.c     | 336 +++++++++++----------------------------
>   drivers/lightnvm/pblk-recovery.c |  69 ++++----
>   drivers/lightnvm/pblk-rl.c       |   3 +-
>   drivers/lightnvm/pblk.h          |  31 ++--
>   include/linux/lightnvm.h         |   2 +
>   10 files changed, 343 insertions(+), 346 deletions(-)
> 

Thanks Igor. I've picked up 5,6, and 8.

Hans will be resend his previous patch to fix the bvec bug. Then we have 
time to get to the right solution without the stress of the -rc window.