On Wed, Feb 9, 2022 at 9:31 AM Dominik Brodowski <linux@xxxxxxxxxxxxxxxxxxxx> wrote: > > Am Wed, Feb 09, 2022 at 02:19:16AM +0100 schrieb Jason A. Donenfeld: > > During crng_init == 0, we never credit entropy in add_interrupt_ > > randomness(), but instead dump it directly into the base_crng. That's > > fine, except for the fact that we then wind up throwing away that > > entropy later when we switch to extracting from the input pool and > > overwriting the base_crng key. The two other early init sites -- > > add_hwgenerator_randomness()'s use crng_fast_load() and add_device_ > > randomness()'s use of crng_slow_load() -- always additionally give their > > inputs to the input pool. But not add_interrupt_randomness(). > > Hm, up to this patch there is no base_crng key. So maybe change the ordering > of the patches? I'll fix the commit message, actually. Eric wrote in his review of v1 that he thinks this problem needs to be fixed before we move to overwriting keys in the subsequent patch and I agreed. Hence, this patch comes first. > > + > > + /* Technically this call means that we're using a spinlock_t > > + * in the IRQ handler, which isn't terrific for PREEMPT_RT. > > + * However, this only happens during very early boot, and then > > Whether it's only during "very early" boot depends on how fast we progress > to crng_init = 2. So maybe just "during boot"? Will do. Jason
