On Wed, 8 May 2019 15:18:10 +0200 (CEST) Sebastian Ott <seb...@linux.ibm.com> wrote:
> > +void cio_gp_dma_free(struct gen_pool *gp_dma, void *cpu_addr, size_t size) > > +{ > > + if (!cpu_addr) > > + return; > > + memset(cpu_addr, 0, size); > > Hm, normally I'd do the memset during alloc not during free - but maybe > this makes more sense here with your usecase in mind. I allocate the backing as zeroed, and zero the memory before putting it back to the pool. So the stuff in the pool is always zeroed. > > > @@ -1063,6 +1163,7 @@ static int __init css_bus_init(void) > > unregister_reboot_notifier(&css_reboot_notifier); > > goto out_unregister; > > } > > + cio_dma_pool_init(); > > This is too late for early devices (ccw console!). You have already raised concern about this last time (thanks). I think, I've addressed this issue: tje cio_dma_pool is only used by the airq stuff. I don't think the ccw console needs it. Please have an other look at patch #6, and explain your concern in more detail if it persists. (@Mimu: What do you think, am I wrong here?) Thanks for having a look! Regards, Halil _______________________________________________ Virtualization mailing list Virtualization@lists.linux-foundation.org https://lists.linuxfoundation.org/mailman/listinfo/virtualization