Re: [PATCH 0/2] Two mmc fixes for IDA and KASAN issues

[Ulf Hansson]

On Wed, 23 Jun 2021 at 09:50, Stephen Boyd <swboyd@xxxxxxxxxxxx> wrote:
>
> Here's a followup to a thread I sent a couple months ago[1]. They're only
> marginally related to each other, but I have bundled them here into one
> series to make it easier to track. Resending to restart the discussion.
>
> Cc: Matthias Schiffer <matthias.schiffer@xxxxxxxxxxxxxxx>
> Cc: Sujit Kautkar <sujitka@xxxxxxxxxxxx>
> Cc: Zubin Mithra <zsm@xxxxxxxxxxxx>
>
> [1] https://lore.kernel.org/r/20210413003621.1403300-1-swboyd@xxxxxxxxxxxx
>
> Stephen Boyd (2):
>   mmc: block: Use kref in place of struct mmc_blk_data::usage
>   mmc: core: Don't allocate IDA for OF aliases
>
>  drivers/mmc/core/block.c | 35 +++++++++++++++++++++--------------
>  drivers/mmc/core/host.c  | 20 ++++++++++----------
>  2 files changed, 31 insertions(+), 24 deletions(-)
>

My apologies for the delay! The changes look very good to me, thanks
for helping out!

FYI, I did some more thorough analysis of how the block device
reference counting/locking should be managed, from a generic point of
view. Your change in patch1 fixes the immediate problem with the KASAN
splat I reported, but also moves the code in the right direction.

However, there are lots of additional improvements that deserve to be
made (another possible KASAN splat) for this related mmc code. I am
looking into this and will keep you posted. :-)

So, this series is applied for fixes and stable tags added for both
patches. Again, thanks for helping out!

Kind regards
Uffe