Re: [RFC PATCH V2] x86, espfix: postpone the initialization of espfix stack for AP

[Ingo Molnar <mingo@kernel.org>]

* Gu Zheng <guz.fnst@cn.fujitsu.com> wrote:

> The following lockdep warning occurrs when running with latest kernel:
> [    3.178000] ------------[ cut here ]------------
> [    3.183000] WARNING: CPU: 128 PID: 0 at kernel/locking/lockdep.c:2755 lockdep_trace_alloc+0xdd/0xe0()
> [    3.193000] DEBUG_LOCKS_WARN_ON(irqs_disabled_flags(flags))
> [    3.199000] Modules linked in:
> 
> [    3.203000] CPU: 128 PID: 0 Comm: swapper/128 Not tainted 4.1.0-rc3 #70
> [    3.221000]  0000000000000000 2d6601fb3e6d4e4c ffff88086fd5fc38 ffffffff81773f0a
> [    3.230000]  0000000000000000 ffff88086fd5fc90 ffff88086fd5fc78 ffffffff8108c85a
> [    3.238000]  ffff88086fd60000 0000000000000092 ffff88086fd60000 00000000000000d0
> [    3.246000] Call Trace:
> [    3.249000]  [<ffffffff81773f0a>] dump_stack+0x4c/0x65
> [    3.255000]  [<ffffffff8108c85a>] warn_slowpath_common+0x8a/0xc0
> [    3.261000]  [<ffffffff8108c8e5>] warn_slowpath_fmt+0x55/0x70
> [    3.268000]  [<ffffffff810ee24d>] lockdep_trace_alloc+0xdd/0xe0
> [    3.274000]  [<ffffffff811cda0d>] __alloc_pages_nodemask+0xad/0xca0
> [    3.281000]  [<ffffffff810ec7ad>] ? __lock_acquire+0xf6d/0x1560
> [    3.288000]  [<ffffffff81219c8a>] alloc_page_interleave+0x3a/0x90
> [    3.295000]  [<ffffffff8121b32d>] alloc_pages_current+0x17d/0x1a0
> [    3.301000]  [<ffffffff811c869e>] ? __get_free_pages+0xe/0x50
> [    3.308000]  [<ffffffff811c869e>] __get_free_pages+0xe/0x50
> [    3.314000]  [<ffffffff8102640b>] init_espfix_ap+0x17b/0x320
> [    3.320000]  [<ffffffff8105c691>] start_secondary+0xf1/0x1f0
> [    3.327000] ---[ end trace 1b3327d9d6a1d62c ]---
> 
> This seems a mis-warning by lockdep, as we alloc pages with GFP_KERNEL in 
> init_espfix_ap() which is called before enabled local irq, and the lockdep 
> sub-system considers this behaviour as allocating memory with GFP_FS with local 
> irq disabled, then trigger the warning as mentioned about.

Why should this be a 'mis-warning'? If the GFP_KERNEL allocation sleeps then we'll 
sleep with irqs disabled => bad.

This looks like a real (albeit hard to trigger) bug.

> Though we could allocate them on the boot CPU side and hand them over to the 
> secondary CPU, but it seemes a bit waste if some of cpus are offline. As thers 
> is no need to these pages(espfix stack) until we try to run user code, so we 
> postpone the initialization of espfix stack after cpu booted to avoid the noise.

> -void init_espfix_ap(void)
> +void init_espfix_ap(int cpu)
>  {

So how about the concern I raised in a former thread, that the allocation should 
be done for the node the target CPU is on? The 'cpu' parameter should be 
propagated to the allocation as well, and turned into a node allocation or so.

Even though some CPUs will share the espfix stack, some won't.

Thanks,

	Ingo