Re: [PATCH] mm: __set_page_dirty_nobuffers uses spin_lock_irqseve instead of spin_lock_irq

2014-02-06 Thread Tang Chen
Hi, Tested-by: Tang Chen Have tested this patch, and the problem is fixed. Thanks. On 02/04/2014 12:49 AM, kosaki.motoh...@gmail.com wrote: From: KOSAKI Motohiro During aio stress test, we observed the following lockdep warning. This mean AIO+numa_balancing is currently deadlockable. The

Re: [PATCH] mm: __set_page_dirty_nobuffers uses spin_lock_irqseve instead of spin_lock_irq

2014-02-06 Thread Tang Chen
Hi, Tested-by: Tang Chen tangc...@cn.fujitsu.com Have tested this patch, and the problem is fixed. Thanks. On 02/04/2014 12:49 AM, kosaki.motoh...@gmail.com wrote: From: KOSAKI Motohirokosaki.motoh...@jp.fujitsu.com During aio stress test, we observed the following lockdep warning. This

Re: [PATCH] mm: __set_page_dirty_nobuffers uses spin_lock_irqseve instead of spin_lock_irq

2014-02-05 Thread Yasuaki Ishimatsu
(2014/02/04 1:49), kosaki.motoh...@gmail.com wrote: From: KOSAKI Motohiro During aio stress test, we observed the following lockdep warning. This mean AIO+numa_balancing is currently deadlockable. The problem is, aio_migratepage disable interrupt, but __set_page_dirty_nobuffers

Re: [PATCH] mm: __set_page_dirty_nobuffers uses spin_lock_irqseve instead of spin_lock_irq

2014-02-05 Thread Yasuaki Ishimatsu
(2014/02/04 1:49), kosaki.motoh...@gmail.com wrote: From: KOSAKI Motohiro kosaki.motoh...@jp.fujitsu.com During aio stress test, we observed the following lockdep warning. This mean AIO+numa_balancing is currently deadlockable. The problem is, aio_migratepage disable interrupt, but

Re: [PATCH] mm: __set_page_dirty_nobuffers uses spin_lock_irqseve instead of spin_lock_irq

2014-02-04 Thread KOSAKI Motohiro
> Indeed, good catch. Do we need the same treatment for > __set_page_dirty_buffers() that can be called by way of > clear_page_dirty_for_io()? Indeed. I posted a patch fixed __set_page_dirty() too. plz see Subject: [PATCH] __set_page_dirty uses spin_lock_irqsave instead of spin_lock_irq -- To

Re: [PATCH] mm: __set_page_dirty_nobuffers uses spin_lock_irqseve instead of spin_lock_irq

2014-02-04 Thread KOSAKI Motohiro
Indeed, good catch. Do we need the same treatment for __set_page_dirty_buffers() that can be called by way of clear_page_dirty_for_io()? Indeed. I posted a patch fixed __set_page_dirty() too. plz see Subject: [PATCH] __set_page_dirty uses spin_lock_irqsave instead of spin_lock_irq -- To

Re: [PATCH] mm: __set_page_dirty_nobuffers uses spin_lock_irqseve instead of spin_lock_irq

2014-02-03 Thread David Rientjes
On Mon, 3 Feb 2014, kosaki.motoh...@gmail.com wrote: > From: KOSAKI Motohiro > > During aio stress test, we observed the following lockdep warning. > This mean AIO+numa_balancing is currently deadlockable. > > The problem is, aio_migratepage disable interrupt, but > __set_page_dirty_nobuffers

[PATCH] mm: __set_page_dirty_nobuffers uses spin_lock_irqseve instead of spin_lock_irq

2014-02-03 Thread kosaki . motohiro
From: KOSAKI Motohiro During aio stress test, we observed the following lockdep warning. This mean AIO+numa_balancing is currently deadlockable. The problem is, aio_migratepage disable interrupt, but __set_page_dirty_nobuffers unintentionally enable it again. Generally, all helper function

[PATCH] mm: __set_page_dirty_nobuffers uses spin_lock_irqseve instead of spin_lock_irq

2014-02-03 Thread kosaki . motohiro
From: KOSAKI Motohiro kosaki.motoh...@jp.fujitsu.com During aio stress test, we observed the following lockdep warning. This mean AIO+numa_balancing is currently deadlockable. The problem is, aio_migratepage disable interrupt, but __set_page_dirty_nobuffers unintentionally enable it again.

Re: [PATCH] mm: __set_page_dirty_nobuffers uses spin_lock_irqseve instead of spin_lock_irq

2014-02-03 Thread David Rientjes
On Mon, 3 Feb 2014, kosaki.motoh...@gmail.com wrote: From: KOSAKI Motohiro kosaki.motoh...@jp.fujitsu.com During aio stress test, we observed the following lockdep warning. This mean AIO+numa_balancing is currently deadlockable. The problem is, aio_migratepage disable interrupt, but