Re: [PATCH] arm/xen: Use alloc_percpu rather than __alloc_percpu

[Stefano Stabellini]

On Wed, 7 Dec 2016, Julien Grall wrote:
> The function xen_guest_init is using __alloc_percpu with an alignment
> which are not power of two.
> 
> However, the percpu allocator never supported alignments which are not power
> of two and has always behaved incorectly in thise case.
> 
> Commit 3ca45a4 "percpu: ensure requested alignment is power of two"
> introduced a check which trigger a warning [1] when booting linux-next
> on Xen. But in fine this bug was always present.
> 
> This can be fixed by replacing the call to __alloc_percpu with
> alloc_percpu. The latter will use an alignment which are a power of two.
> 
> [1]
> 
> [    0.023921] illegal size (48) or align (48) for percpu allocation
> [    0.024167] ------------[ cut here ]------------
> [    0.024344] WARNING: CPU: 0 PID: 1 at linux/mm/percpu.c:892 pcpu_alloc+0x88/0x6c0
> [    0.024584] Modules linked in:
> [    0.024708]
> [    0.024804] CPU: 0 PID: 1 Comm: swapper/0 Not tainted
> 4.9.0-rc7-next-20161128 #473
> [    0.025012] Hardware name: Foundation-v8A (DT)
> [    0.025162] task: ffff80003d870000 task.stack: ffff80003d844000
> [    0.025351] PC is at pcpu_alloc+0x88/0x6c0
> [    0.025490] LR is at pcpu_alloc+0x88/0x6c0
> [    0.025624] pc : [<ffff00000818e678>] lr : [<ffff00000818e678>]
> pstate: 60000045
> [    0.025830] sp : ffff80003d847cd0
> [    0.025946] x29: ffff80003d847cd0 x28: 0000000000000000
> [    0.026147] x27: 0000000000000000 x26: 0000000000000000
> [    0.026348] x25: 0000000000000000 x24: 0000000000000000
> [    0.026549] x23: 0000000000000000 x22: 00000000024000c0
> [    0.026752] x21: ffff000008e97000 x20: 0000000000000000
> [    0.026953] x19: 0000000000000030 x18: 0000000000000010
> [    0.027155] x17: 0000000000000a3f x16: 00000000deadbeef
> [    0.027357] x15: 0000000000000006 x14: ffff000088f79c3f
> [    0.027573] x13: ffff000008f79c4d x12: 0000000000000041
> [    0.027782] x11: 0000000000000006 x10: 0000000000000042
> [    0.027995] x9 : ffff80003d847a40 x8 : 6f697461636f6c6c
> [    0.028208] x7 : 6120757063726570 x6 : ffff000008f79c84
> [    0.028419] x5 : 0000000000000005 x4 : 0000000000000000
> [    0.028628] x3 : 0000000000000000 x2 : 000000000000017f
> [    0.028840] x1 : ffff80003d870000 x0 : 0000000000000035
> [    0.029056]
> [    0.029152] ---[ end trace 0000000000000000 ]---
> [    0.029297] Call trace:
> [    0.029403] Exception stack(0xffff80003d847b00 to
>                                0xffff80003d847c30)
> [    0.029621] 7b00: 0000000000000030 0001000000000000
> ffff80003d847cd0 ffff00000818e678
> [    0.029901] 7b20: 0000000000000002 0000000000000004
> ffff000008f7c060 0000000000000035
> [    0.030153] 7b40: ffff000008f79000 ffff000008c4cd88
> ffff80003d847bf0 ffff000008101778
> [    0.030402] 7b60: 0000000000000030 0000000000000000
> ffff000008e97000 00000000024000c0
> [    0.030647] 7b80: 0000000000000000 0000000000000000
> 0000000000000000 0000000000000000
> [    0.030895] 7ba0: 0000000000000035 ffff80003d870000
> 000000000000017f 0000000000000000
> [    0.031144] 7bc0: 0000000000000000 0000000000000005
> ffff000008f79c84 6120757063726570
> [    0.031394] 7be0: 6f697461636f6c6c ffff80003d847a40
> 0000000000000042 0000000000000006
> [    0.031643] 7c00: 0000000000000041 ffff000008f79c4d
> ffff000088f79c3f 0000000000000006
> [    0.031877] 7c20: 00000000deadbeef 0000000000000a3f
> [    0.032051] [<ffff00000818e678>] pcpu_alloc+0x88/0x6c0
> [    0.032229] [<ffff00000818ece8>] __alloc_percpu+0x18/0x20
> [    0.032409] [<ffff000008d9606c>] xen_guest_init+0x174/0x2f4
> [    0.032591] [<ffff0000080830f8>] do_one_initcall+0x38/0x130
> [    0.032783] [<ffff000008d90c34>] kernel_init_freeable+0xe0/0x248
> [    0.032995] [<ffff00000899a890>] kernel_init+0x10/0x100
> [    0.033172] [<ffff000008082ec0>] ret_from_fork+0x10/0x50
> 
> Reported-by: Wei Chen <wei.chen@xxxxxxx>
> Link: https://lkml.org/lkml/2016/11/28/669
> Signed-off-by: Julien Grall <julien.grall@xxxxxxx>
> 
> Cc: stable@xxxxxxxxxxxxxxx
> 
> ---
> 
> I have requested backport to stable because the percpu allocator may
> misbehaves when the alignment is not a power of two.

Reviewed-by: Stefano Stabellini <sstabellini@xxxxxxxxxx>

I'll fix a small wording issue in the commit message and apply to
xentip.


>  arch/arm/xen/enlighten.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/arch/arm/xen/enlighten.c b/arch/arm/xen/enlighten.c
> index f193414..4986dc0 100644
> --- a/arch/arm/xen/enlighten.c
> +++ b/arch/arm/xen/enlighten.c
> @@ -372,8 +372,7 @@ static int __init xen_guest_init(void)
>  	 * for secondary CPUs as they are brought up.
>  	 * For uniformity we use VCPUOP_register_vcpu_info even on cpu0.
>  	 */
> -	xen_vcpu_info = __alloc_percpu(sizeof(struct vcpu_info),
> -			                       sizeof(struct vcpu_info));
> +	xen_vcpu_info = alloc_percpu(struct vcpu_info);
>  	if (xen_vcpu_info == NULL)
>  		return -ENOMEM;
>  
> -- 
> 1.9.1
>