Re: scheduling in kmalloc()

[Alan Cox]

> if we use any thing other than GFP_ATOMIC, does it result in scheduling
> out the process if there is no memory available?
> with GFP_KERNRL, I think we try freeing pages to service the current
> request.
> or is there any possibility of kmalloc() failing even with GFP_KERNEL?

kmalloc can always fail, looping on a kmalloc at high level can almost always
cause deadlocks so you need to be prepared to fail