xfs
[Top] [All Lists]

Re: [PATCH] mm/vmscan: Do not block forever at shrink_inactive_list().

To: Tetsuo Handa <penguin-kernel@xxxxxxxxxxxxxxxxxxx>
Subject: Re: [PATCH] mm/vmscan: Do not block forever at shrink_inactive_list().
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Tue, 20 May 2014 15:24:52 +1000
Cc: riel@xxxxxxxxxx, kosaki.motohiro@xxxxxxxxxxxxxx, fengguang.wu@xxxxxxxxx, kamezawa.hiroyu@xxxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <201405200354.s4K3sTUx057458@xxxxxxxxxxxxxxxxxxx>
References: <201405192340.FCD48964.OFQHOOJLVSFFMt@xxxxxxxxxxxxxxxxxxx> <20140520004449.GE18954@dastard> <201405200354.s4K3sTUx057458@xxxxxxxxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Tue, May 20, 2014 at 12:54:29PM +0900, Tetsuo Handa wrote:
> Dave Chinner wrote:
> > So, XFS should be passing kswapd context to the workqueue allocation
> > context. The patch below does this.
> > 
> > Tetsuo-san, when it comes to problems involving XFS, you should
> > really CC xfs@xxxxxxxxxxx because very few people really know how
> > XFS works and even fewer still know how it is supposed to interact
> > with memory reclaim....
> 
> Thank you for the patch, but ...
> 
>   #define PF_KSWAPD       0x00040000      /* I am kswapd */
> 
>   static inline int current_is_kswapd(void)
>   {
>       return current->flags & PF_KSWAPD;
>   }

> I think ((char) (current->flags & 0x00040000)) == 0.
> Your patch wants
> 
>   -args->kswapd = current_is_kswapd();
>   +args->kswapd = (current_is_kswapd() != 0);

Thanks for pointing that out, but I think:

-static inline int current_is_kswapd(void)
+static inline bool current_is_kswapd(void)

is a better solution. It can only be true or false.

But regardless, I need to change the boolean options in that XFS
structure to be, well, booleans.

Cheers,

dave.

-- 
Dave Chinner
david@xxxxxxxxxxxxx

<Prev in Thread] Current Thread [Next in Thread>