xfs
[Top] [All Lists]

Re: [PATCH 2/4] xfs: replace i_flock with a sleeping bitlock

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: [PATCH 2/4] xfs: replace i_flock with a sleeping bitlock
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Wed, 19 Oct 2011 11:42:06 +1100
Cc: xfs@xxxxxxxxxxx
In-reply-to: <20111018201405.357001594@xxxxxxxxxxxxxxxxxxxxxx>
References: <20111018201304.279051318@xxxxxxxxxxxxxxxxxxxxxx> <20111018201405.357001594@xxxxxxxxxxxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Tue, Oct 18, 2011 at 04:13:06PM -0400, Christoph Hellwig wrote:
> We almost never block on i_flock, the exception is synchronous inode
> flushing.  Instead of bloating the inode with a 16/24-byte completion
> that we abuse as a semaphore just implement it as a bitlock that uses
> a bit waitqueue for the rare sleeping path.  This primarily is a
> tradeoff between a much smaller inode and a faster non-blocking
> path vs a faster faster wakeups, and we are much better off with
> the former.
> 
> A small downside is that we will lose lockdep checking for i_flock, but
> given that it's always taken inside the ilock that should be acceptable.
> 
> Note that for example the inode writeback locking is implemented in a
> very similar way.
> 
> Signed-off-by: Christoph Hellwig <hch@xxxxxx>

.....
> @@ -716,3 +716,19 @@ xfs_isilocked(
>       return 0;
>  }
>  #endif
> +
> +void
> +__xfs_iflock(
> +     struct xfs_inode        *ip)
> +{
> +     wait_queue_head_t *wq = bit_waitqueue(&ip->i_flags, __XFS_IFLOCK);
> +     DEFINE_WAIT_BIT(wait, &ip->i_flags, __XFS_IFLOCK);
> +
> +     do {
> +             prepare_to_wait_exclusive(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
> +             if (xfs_isiflocked(ip))
> +                     schedule();
> +     } while (!xfs_iflock_nowait(ip));
> +
> +     finish_wait(wq, &wait.wait);
> +}

Given that the only way that the inode will become unlocked is for
IO to complete, that makes this an IO wait, right? Perhaps this
should call io_schedule() in that case?

> @@ -380,6 +372,8 @@ static inline void xfs_ifunlock(xfs_inod
>  #define XFS_IFILESTREAM              0x0010  /* inode is in a filestream 
> directory */
>  #define XFS_ITRUNCATED               0x0020  /* truncated down so 
> flush-on-close */
>  #define XFS_IDIRTY_RELEASE   0x0040  /* dirty release already seen */
> +#define __XFS_IFLOCK         8       /* inode is beeing flushed right now */
> +#define XFS_IFLOCK           (1 << __XFS_IFLOCK)

Any reason for leaving a gap in the flag space here?

Otherwise looks good.

Reviewed-by: Dave Chinner <dchinner@xxxxxxxxxx>

-- 
Dave Chinner
david@xxxxxxxxxxxxx

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