On Wed, Jan 24, 2024 at 08:44:16AM -0700, Keith Busch wrote: > > > > + min_not_zero(q->limits.max_hw_discard_sectors, > > > > + q->limits.max_user_discard_sectors); > > > > > > Shouldn't writing 0 disable discards? > > > > I mirror the max_user_sectors behavior here, where 0 disables the > > user limiting. But yes, that would be a behavior change. > > But a user should be able to disable discards, right? Unless you really > want to match max_user_sectors, I think you could default the user > setting here to UINT_MAX and use "min" instead of "min_not_zero". Not a fan of having different semantics for the different attributes, but given that we've had this or 9 years we better stick to it, so I'll change it.
