On Tue, 20 Jun 2023 at 12:19, Linus Walleij <linus.walleij@xxxxxxxxxx> wrote: > > On Tue, Jun 20, 2023 at 11:27 AM Yann Gautier <yann.gautier@xxxxxxxxxxx> wrote: > > > STM32 variants also have busy_detect = true. > > Could that be an issue to initialize this work, which seem dedicated to > > ux500? > > As Ulf says it is just initialized then it is left unused. > > I renamed the software timeout ux500_busy_timeout_work() from the > previous name busy_timeout_work() but I actually thought it could > make sense to enable it on STM32 as well. It is essentially a last > resort fallback if the card does not properly report busy end or the block > fails to detect it for some reason. There is HW busy-timeout for the STM32 variant, so I don't think it's needed. However, that HW timeout does have an upper limit. If the requested timeout for the command is longer than what the HW can support, maybe a software timeout can be used in some clever way. The tricky part is, if I recall correctly from previous conversations at LKML, one can not skip to set the HW timeout. This means, this is going to be a bit more messy to support. Br Uffe
