| To: | Steve Lord <lord@xxxxxxx> |
|---|---|
| Subject: | Re: TAKE - implement direct read and write in XFS |
| From: | Andi Kleen <ak@xxxxxxx> |
| Date: | Mon, 5 Mar 2001 18:08:44 +0100 |
| Cc: | linux-xfs@xxxxxxxxxxx |
| In-reply-to: | <200103051654.f25GsbI16800@xxxxxxxxxxxxxxxxxxxx>; from lord@xxxxxxx on Mon, Mar 05, 2001 at 10:54:37AM -0600 |
| References: | <200103051654.f25GsbI16800@xxxxxxxxxxxxxxxxxxxx> |
| Sender: | owner-linux-xfs@xxxxxxxxxxx |
| User-agent: | Mutt/1.2.5i |
On Mon, Mar 05, 2001 at 10:54:37AM -0600, Steve Lord wrote: > This implements direct I/O for xfs on linux, read and write should function. > Testing has been limited so far. There is cache coherency between buffered > and direct I/O to the extent that the direct calls will flush or flush and > invalidate the cache before starting I/O. A systune has been added at > /proc/sys/vm/pagebuf/max_dio_pages which defines how many pages will be > locked down at once during an individual read or write call, this does not > limit the size of the I/O, just the size of the requests which will be > issued to the block layer. Very cool :-) -Andi |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Getting started..., Joshua Baker-LePain |
|---|---|
| Next by Date: | Re: questions about some benchmarks, Steve Lord |
| Previous by Thread: | TAKE - implement direct read and write in XFS, Steve Lord |
| Next by Thread: | Getting started..., Joshua Baker-LePain |
| Indexes: | [Date] [Thread] [Top] [All Lists] |