Re: MTD: Patch to spi-nor hangs Lenovo ThinkPad X1 on suspend, sleep, and shutdown.

[Todd Brandt]

I just sent out a fix patch for 6.4-rc2, let's move the discussion to
that mail:

[PATCH] MTD SPI-NOR: BUG FIX of divide by zero in new n_banks value
https://marc.info/?l=linux-kernel&m=168427739123254&w=2

On Tue, 2023-05-16 at 12:12 -0700, Todd Brandt wrote:
> This is the kernel bugzilla: 
> https://bugzilla.kernel.org/show_bug.cgi?id=217448
> 
> This is the system I'm having troubles with:
> 
> os-version              : Ubuntu 20.04.2 LTS
> baseboard-manufacturer  : LENOVO
> baseboard-product-name  : 20QA000FUS
> baseboard-serial-number : W1KS11R111D
> baseboard-version       : SDK0J40697 WIN
> bios-release-date       : 05/20/2021
> bios-vendor             : LENOVO
> bios-version            : N2MET49W (1.14 )
> chassis-manufacturer    : LENOVO
> chassis-serial-number   : R911Q4MM
> chassis-version         : None
> processor-manufacturer  : Intel(R) Corporation
> processor-version       : 11th Gen Intel(R) Core(TM) i7-1160G7 @
> 1.20GHz
> system-manufacturer     : LENOVO
> system-product-name     : 20QA000FUS
> system-serial-number    : R911Q4MM
> system-version          : ThinkPad X1 Titanium Gen 1
> cpucount                : 8
> memtotal                : 15939728 kB
> memfree                 : 8428028 kB
> 
> Ever since 6.4-rc1 this system has hung on every S3/S2idle suspend
> and
> shutdown during reboot. I've bisected it to a patch to the MTD: SPI-
> NOR 
> subsystem about introducing the "concept of a bank"
> 
> 9d6c5d64f0288a814d4435b7da39e360a4c39e40 is the first bad commit
> commit 9d6c5d64f0288a814d4435b7da39e360a4c39e40
> Author: Miquel Raynal <miquel.raynal@xxxxxxxxxxx>
> Date:   Tue Mar 28 17:40:58 2023 +0200
> 
>     mtd: spi-nor: Introduce the concept of bank
> 
>     SPI NOR chips are made of pages, which gathered in small groups
> make
>     (erase) sectors. Sectors, gathered together, make banks inside
> the
>     chip. Until now, there was only one bank per device supported,
> but
> we
>     are about to introduce support for new chips featuring several
> banks (up
>     to 4 so far) where different operations may happen in parallel.
> 
>     Let's allow describing these additional bank parameters, and
> let's
> do
>     this independently of any other value (like the number of
> sectors)
> with
>     an absolute value.
> 
>     By default we consider that all chips have a single bank.
> 
>     Signed-off-by: Miquel Raynal <miquel.raynal@xxxxxxxxxxx>
>     Reviewed-by: Pratyush Yadav <pratyush@xxxxxxxxxx>
>     Link: 
> 
https://lore.kernel.org/r/20230328154105.448540-2-miquel.raynal@xxxxxxxxxxx
>     Signed-off-by: Tudor Ambarus <tudor.ambarus@xxxxxxxxxx>
> 
>  drivers/mtd/spi-nor/core.c   |  1 +
>  drivers/mtd/spi-nor/core.h   | 16 +++++++++++-----
>  drivers/mtd/spi-nor/xilinx.c |  1 +
>  3 files changed, 13 insertions(+), 5 deletions(-)
> 
> Basically something about the new code is improperly handling the
> memory size on this machine. It's not complex but I haven't dug any
> deeper since I want it posted as quickly as possible. To reproduce,
> run
> any of these 3 commands and the system will hang:
> 
> echo freeze > /sys/power/state
> echo mem > /sys/power/state
> sudo reboot