Re: [PATCH v4 0/4] mtd: core: protect access to mtd devices while in suspend

[Boris Brezillon]

On Tue, 26 Oct 2021 07:55:47 +0200
Sean Nyekjaer <sean@xxxxxxxxxx> wrote:

> Follow-up on discussion in:
> https://lkml.org/lkml/2021/10/4/41
> https://lkml.org/lkml/2021/10/11/435
> 
> Changes since v3:
>  - edited commit msg and author for mtdconcat patch
> 
> Changes since v2:
>  - added signoff's to patch from Boris
>  - removed accidential line break
>  - kept tests consistent: master->master.suspended == 0 -> !master->master.suspended
>  - added comments to mtdconcat patch
>  - moved mtdconcat before ('mtd: core: protect access to MTD devices while in suspend')
> 
> Changes since v1:
>  - removed __mtd_suspend/__mtd_resume functions as they are not used by
>    mtdconcat anymore.
>  - only master mtd_info is used for mtd_{start,end}_access(). Warn if we
>    got mtd's.
>  - added Boris patch for using uninitialized _suspend/_resume hooks when
>    bbt scanning
>  - mtdconcat uses device _suspend/_resume hooks
>  - I don't really like the macro proposal from Boris
>    mtd_no_suspend_void_call()/mtd_no_suspend_ret_call() I think they
>    make the code complex to read and the macro's doesn't fit every
>    where anyway...
> 
> Changes since from rfc v1/v2:
>  - added access protection for all device access hooks in mtd_info.
>  - added Suggested-by to [1/3] patch.
>  - removed refereces to commit ef347c0cfd61 ("mtd: rawnand: gpmi: Implement exec_op")
>    from commit msg as commit 013e6292aaf5 ("mtd: rawnand: Simplify the locking") is 
>    to be blamed.
>  - tested on a kernel with LOCKDEP enabled.
> 
> Boris Brezillon (2):
>   mtd: rawnand: nand_bbt: hide suspend/resume hooks while scanning bbt
>   mtd: mtdconcat: don't use mtd_{suspend,resume}()
> 
> Sean Nyekjaer (2):
>   mtd: core: protect access to MTD devices while in suspend
>   mtd: rawnand: remove suspended check

Looks good overall (after fixing the minor things I pointed out, of
course), but I'd recommend applying this series to mtd-next just after
-rc1 is out so you get a chance to detect regressions before it's
merged in Linus' tree. I fear this unconditional blocking on suspended
device will lead to unexpected deadlocks (see my comment on panic
writes)...

> 
>  drivers/mtd/mtdconcat.c          |  15 +++-
>  drivers/mtd/mtdcore.c            | 124 +++++++++++++++++++++++++++----
>  drivers/mtd/nand/raw/nand_base.c |  52 ++++---------
>  drivers/mtd/nand/raw/nand_bbt.c  |  28 ++++++-
>  include/linux/mtd/mtd.h          |  81 ++++++++++++++++----
>  include/linux/mtd/rawnand.h      |   5 +-
>  6 files changed, 230 insertions(+), 75 deletions(-)
>