On Sat, Mar 28, 2020 at 07:56:17PM -0700, Matthew Wilcox wrote: > On Fri, Mar 27, 2020 at 06:11:47PM +1100, Alastair D'Silva wrote: > > +static struct mutex minors_idr_lock; > > +static struct idr minors_idr; > ... > > + mutex_lock(&minors_idr_lock); > > + minor = idr_alloc(&minors_idr, ocxlpmem, 0, NUM_MINORS, GFP_KERNEL); > > + mutex_unlock(&minors_idr_lock); > ... > > + mutex_lock(&minors_idr_lock); > > + idr_remove(&minors_idr, MINOR(ocxlpmem->dev.devt)); > > + mutex_unlock(&minors_idr_lock); > ... > > + mutex_init(&minors_idr_lock); > > + idr_init(&minors_idr); > > Unless you look up ocxlpmem by minor number later in the patch series (and > most of the series didn't make it to my mailbox), this can just be an ida. > > static DEFINE_IDA(minors); > ... > minor = ida_alloc_max(&minors, NUM_MINORS, GFP_KERNEL); Oops. NUM_MINORS - 1. Or #define MAX_MINOR 255 since you only seem to use NUM_MINORS in this one place. > ... > ida_free(&minors, MINOR(ocxlpmem->dev.devt)); > ... > and you can drop the dynamic initialisation. And the mutex.
