On 6/19/24 11:52 AM, Sebastian Andrzej Siewior wrote: > On 2024-06-19 11:34:23 [-0600], Jens Axboe wrote: >> On 6/19/24 9:08 AM, Sebastian Andrzej Siewior wrote: >>> From: Mike Galbraith <umgwanakikbuti@xxxxxxxxx> >>> >>> The bit spinlock disables preemption. The spinlock_t lock becomes a sleeping >>> lock on PREEMPT_RT and it can not be acquired in this context. In this locked >>> section, zs_free() acquires a zs_pool::lock, and there is access to >>> zram::wb_limit_lock. >>> >>> Use a spinlock_t on PREEMPT_RT for locking and set/ clear ZRAM_LOCK bit after >>> the lock has been acquired/ dropped. >> >> The conditional code depending on CONFIG_PREEMPT_RT is nasty. Why not >> just get rid of that and use the CONFIG_PREEMPT_RT variants for >> everything? They are either good enough to work well in general, or it >> should be redone such that it is. > > That would increase the struct size with lockdep for !RT. But it is > probably not a concern. Also other bits (besides ZRAM_LOCK) can not be > added but that wasn't needed in the last few years. Yeah I really don't think anyone cares about the struct size when PROVE_LOCKING is on... -- Jens Axboe
