[Top] [All Lists]

Re: Question about GFP_KERNEL argument in radix_tree_preload

To: Sanidhya Kashyap <sanidhya.gatech@xxxxxxxxx>
Subject: Re: Question about GFP_KERNEL argument in radix_tree_preload
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Mon, 23 Mar 2015 15:28:47 +1100
Cc: changwoo@xxxxxxxxxx, "Kashyap, Sanidhya" <sanidhya@xxxxxxxxxx>, taesoo@xxxxxxxxxx, blee@xxxxxxxxxx, csong@xxxxxxxxxx, xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <CADa969g-jGyYc1ts+JamuD0Ze-Mcp+X8am7rAfDeY0wXC4A6gg@xxxxxxxxxxxxxx>
References: <CADa969g-jGyYc1ts+JamuD0Ze-Mcp+X8am7rAfDeY0wXC4A6gg@xxxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Sun, Mar 22, 2015 at 11:49:34PM -0400, Sanidhya Kashyap wrote:
> Hello David,
> We have been doing some analysis on file systems and have a question
> regarding the GFP_KERNEL flag used in radix_tree_preload in
> xfs_mru_cache_insert.
> We think that this should be GFP_NOFS as it is related to the block
> allocation and other file systems(eg btrfs,ext2) are also using
> GFP_NOFS for the same.
> It would be great if you can let us know your reason for using that
> flag and which one is the correct option?


That code was written by someone else and it was 7-8 years ago I
cleaned it up for Linux. Whatever reason there was is long since

If it's wrong (e.g. can be called inside a transaction), send a
patch to fix it.


Dave Chinner

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