On Wed, 2010-03-31 at 14:28 -0700, Andrew Morton wrote: > > Well, it's all a bit nasty. kmem_cache_create() does a lot of stuff, > including calling into the page allocator with GFP_KERNEL - expecting > kmem_cache_create() to preserve local_irq_disable() is a bit > optimistic. Well, the sl*b allocator -has- been modified to avoid enabling IRQs early, at least I remember we did that when we moved it to be initialized earlier. > radix_tree_init() calls hotcpu_notifier() which also does > mutex_lock(&cpu_add_remove_lock); > > The easiest fix is to reposition the interrutps-are-now-enabled point > in start_kernel(). But I have a feeling that some versions of > early_irq_init() won't like that. Yeah that won't work. Interrupts must not be enabled before at least init_IRQ() and time_init(). The problem is that until all these guys have gone through their initializations, there may be pending spurrious crap coming from the HW (timers, external IRQs, profile IRQs) due to such HW not yet properly "sanitized" by the kernel. Plenty of archs have those assumptions wired in. I don't think moving the IRQ enable point earlier is the right approach. Cheers, Ben. -- To unsubscribe from this list: send the line "unsubscribe linux-arch" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html