| To: | Chris Wedgwood <cw@xxxxxxxx> |
|---|---|
| Subject: | Re: Here is the results of my last question... |
| From: | Russell Cattelan <cattelan@xxxxxxxxxxx> |
| Date: | 28 Oct 2002 09:51:08 -0600 |
| Cc: | Andi Kleen <ak@xxxxxxx>, yoros@xxxxxxxxxx, linux-xfs <linux-xfs@xxxxxxxxxxx> |
| In-reply-to: | <20021025194911.GB26998@xxxxxxxxxxxxx> |
| References: | <20021025181514.GA21940@xxxxxxxxxxxxxxxxxxx> <20021025202727.A2421@xxxxxxxxxxxxx> <20021025194911.GB26998@xxxxxxxxxxxxx> |
| Sender: | linux-xfs-bounce@xxxxxxxxxxx |
On Fri, 2002-10-25 at 14:49, Chris Wedgwood wrote: > On Fri, Oct 25, 2002 at 08:27:27PM +0200, Andi Kleen wrote: > > > The right thing would be to change the downloader assistant to use > > XFS' file preallocation ioctls so that it allocates the full space > > in advance. Unfortuately they are not fully implemented on linux/xfs > > and are only available for root currently. I would be possible to > > create the file first and fill it with zeroes. This way you should > > get few extents. > > It would be pretty cheap for the downloader to actually just touch > lseek/touch ahead a few MB of each writer too, this would be almost as > easy to code and would work on other filesystems. That would just create a file with a hole, and holes are filled in the same way appends are filled in. > > > --cw > |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: Reason for delay in inclusion of ACLs to mainstream kernel?, Steve Lord |
|---|---|
| Next by Date: | Re: kernel-2.4.19-SGI_XFS_1.2pre2.src.rpm?, Russell Cattelan |
| Previous by Thread: | Re: Here is the results of my last question..., Chris Wedgwood |
| Next by Thread: | Re: Here is the results of my last question..., Chris Wedgwood |
| Indexes: | [Date] [Thread] [Top] [All Lists] |