From: Mingming Cao <cmm@us.ibm.com>
To: Mingming Cao <cmm@us.ibm.com>
Cc: Ray Lee <ray-lk@madrabbit.org>, Andrew Morton <akpm@osdl.org>,
sct@redhat.com, pbadari@us.ibm.com,
Linux Kernel <linux-kernel@vger.kernel.org>,
ext2-devel@lists.sourceforge.net
Subject: Re: [Ext2-devel] Re: [PATCH 2/3] ext3 reservation allow turn off for specifed file
Date: 25 Oct 2004 16:05:44 -0700 [thread overview]
Message-ID: <1098745548.9754.7427.camel@w-ming2.beaverton.ibm.com> (raw)
In-Reply-To: <1098736389.9692.7243.camel@w-ming2.beaverton.ibm.com>
On Mon, 2004-10-25 at 13:33, Mingming Cao wrote:
> On Wed, 2004-10-20 at 10:55, Ray Lee wrote:
> > On Mon, 2004-10-18 at 18:01 -0700, Mingming Cao wrote:
> > > On Mon, 2004-10-18 at 16:42, Andrew Morton wrote:
> > > > Applications currently pass a seeky-access hint into the kernel via
> > > > posix_fadvise(POSIX_FADV_RANDOM). It would be nice to hook into that
> >
> > [...]
> >
> > > Just thought seeky random write application could use the existing ioctl
> > > to let the kernel know it does not need reservation at all. Isn't that
> > > more straightforward?
> >
> > Going the ioctl route seems to imply that userspace would have to do a
> > posix_fadvise() call and the ioctl, as opposed to just the fadvise. No?
> > I'm betting the fadvise call is a little more portable as well.
>
> Agreed. How about this: add a check in ext3_prepare_write(), if user passed seeky-access hint through posix_fadvise(via check for file->f_ra.ra_pages == 0), if so, close the reservation window. But we still need previous patch to handle window size 0(no reservation) in reservation code.
The previous patch could re-open the reservation in the case user then
switch back to POSIX_FADV_NORMAL or POSIZ_FADV_SEQUENTAIL. Updated
version included. We set the window size to -1 if the user pass the
POSIX_FADV_RANDOM, instead of 0, to differentiate the case when user
really want to close the window by ioctl. We could re-open the window
only when user pass the POSIX_FADV_SEQUENTAIL (or POSIX_FADV_NORMAL
hint) and the window was closed because of previous seeky access hint.
Currently the readahead is turned off if the userspace passed a seeky access hint to kernel by POSIX_FADVISE. It would be nice to turn off the block allocation reservation as well for seeky random write.
---
linux-2.6.9-rc4-mm1-ming/fs/ext3/balloc.c | 2 +-
linux-2.6.9-rc4-mm1-ming/fs/ext3/inode.c | 17 +++++++++++++++++
2 files changed, 18 insertions(+), 1 deletion(-)
diff -puN fs/ext3/inode.c~ext3_turn_off_reservation_by_fadvise fs/ext3/inode.c
--- linux-2.6.9-rc4-mm1/fs/ext3/inode.c~ext3_turn_off_reservation_by_fadvise 2004-10-25 18:25:44.135751800 -0700
+++ linux-2.6.9-rc4-mm1-ming/fs/ext3/inode.c 2004-10-25 21:43:16.194964928 -0700
@@ -1001,6 +1001,7 @@ static int ext3_prepare_write(struct fil
int ret, needed_blocks = ext3_writepage_trans_blocks(inode);
handle_t *handle;
int retries = 0;
+ int windowsz = 0;
retry:
handle = ext3_journal_start(inode, needed_blocks);
@@ -1008,6 +1009,22 @@ retry:
ret = PTR_ERR(handle);
goto out;
}
+
+ /*
+ * if user passed a seeky-access hint to kernel,
+ * through POSIX_FADV_RANDOM,(file->r_ra.ra_pages is cleared)
+ * turn off reservation for block allocation correspondingly.
+ *
+ * Otherwise, if user switch back to POSIX_FADV_SEQUENTIAL or
+ * POSIX_FADV_NORMAL, re-open the reservation window.
+ */
+ windowsz = atomic_read(&EXT3_I(inode)->i_rsv_window.rsv_goal_size);
+ if ((windowsz > 0) && (!file->f_ra.ra_pages))
+ atomic_set(&EXT3_I(inode)->i_rsv_window.rsv_goal_size, -1);
+ if ((windowsz == -1) && file->f_ra.ra_pages)
+ atomic_set(&EXT3_I(inode)->i_rsv_window.rsv_goal_size,
+ EXT3_DEFAULT_RESERVE_BLOCKS);
+
ret = block_prepare_write(page, from, to, ext3_get_block);
if (ret)
goto prepare_write_failed;
diff -puN fs/ext3/balloc.c~ext3_turn_off_reservation_by_fadvise fs/ext3/balloc.c
--- linux-2.6.9-rc4-mm1/fs/ext3/balloc.c~ext3_turn_off_reservation_by_fadvise 2004-10-25 21:23:05.152071432 -0700
+++ linux-2.6.9-rc4-mm1-ming/fs/ext3/balloc.c 2004-10-25 21:24:48.136415432 -0700
@@ -1154,7 +1154,7 @@ int ext3_new_block(handle_t *handle, str
struct ext3_sb_info *sbi;
struct reserve_window_node *my_rsv = NULL;
struct reserve_window_node *rsv = &EXT3_I(inode)->i_rsv_window;
- unsigned short windowsz = 0;
+ int windowsz = 0;
#ifdef EXT3FS_DEBUG
static int goal_hits, goal_attempts;
#endif
_
next prev parent reply other threads:[~2004-10-25 23:36 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2004-10-15 13:27 Ext3 -mm reservations code: is this fix really correct? Stephen C. Tweedie
2004-10-15 16:01 ` [Ext2-devel] " mingming cao
2004-10-15 16:40 ` Stephen C. Tweedie
2004-10-15 20:29 ` mingming cao
2004-10-15 22:20 ` Stephen C. Tweedie
2004-10-15 22:34 ` mingming cao
2004-10-18 22:55 ` [PATCH 2/3] ext3 reservation allow turn off for specifed file Mingming Cao
2004-10-18 23:42 ` Andrew Morton
2004-10-19 1:01 ` Mingming Cao
2004-10-20 17:55 ` Ray Lee
2004-10-25 20:33 ` [Ext2-devel] " Mingming Cao
2004-10-25 23:05 ` Mingming Cao [this message]
2004-10-25 23:45 ` Andrew Morton
2004-10-26 16:53 ` Mingming Cao
2004-10-26 20:18 ` Andrew Morton
2004-10-26 23:01 ` Mingming Cao
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1098745548.9754.7427.camel@w-ming2.beaverton.ibm.com \
--to=cmm@us.ibm.com \
--cc=akpm@osdl.org \
--cc=ext2-devel@lists.sourceforge.net \
--cc=linux-kernel@vger.kernel.org \
--cc=pbadari@us.ibm.com \
--cc=ray-lk@madrabbit.org \
--cc=sct@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®