On Mon, Nov 02, 2015 at 12:11:33PM -0600, Christoph Lameter wrote: > On Mon, 2 Nov 2015, Catalin Marinas wrote: > > > On Mon, Nov 02, 2015 at 10:48:17AM -0600, Christoph Lameter wrote: > > > On Mon, 2 Nov 2015, Catalin Marinas wrote: > > > > I haven't looked at the patch 3/3 in detail but I'm pretty sure I'll NAK > > > > the approach (and the definition of PERCPU_ATOM_SIZE), therefore > > > > rendering this patch unnecessary. IIUC, this is used to enforce some > > > > alignment of the per-CPU IRQ stack to be able to check whether the > > > > current stack is process or IRQ on exception entry. But there are other, > > > > less intrusive ways to achieve the same (e.g. x86). > > > > > > The percpu allocator allows the specification of alignment requirements. > > > > Patch 3/3 does something like this: > > > > DEFINE_PER_CPU(char [IRQ_STACK_SIZE], irq_stacks) __aligned(IRQ_STACK_SIZE) > > > > where IRQ_STACK_SIZE > PAGE_SIZE. AFAICT, setup_per_cpu_areas() doesn't > > guarantee alignment greater than PAGE_SIZE. > > And we cannot use percpu_alloc() instead? Aligning the whole of the percpu > area because one allocation requires it? I haven't tried but it seems that pcpu_alloc() has a WARN() when align > PAGE_SIZE and it would fail. As I said in a previous reply, I don't think this patch is necessary, mainly because I don't particularly like the logic for detecting the IRQ stack re-entrance based on the stack pointer alignment. -- Catalin -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>