Re: [RESEND][PATCH v2] block: remove plugging at buffered write time

From: Fengguang Wu
Date: Wed Apr 11 2012 - 21:32:26 EST


On Wed, Apr 11, 2012 at 04:13:44PM -0700, Andrew Morton wrote:
> On Mon, 09 Apr 2012 10:34:44 -0400
> Jeff Moyer <jmoyer@xxxxxxxxxx> wrote:
>
> > Wu Fengguang <fengguang.wu@xxxxxxxxx> writes:
> >
> > > Buffered write(2) is not directly tied to IO, so it's not suitable to
> > > handle plug in generic_file_aio_write().
> > [snip, moved]
> > > Note that plugging for O_SYNC writes is also removed. The user may pass
> > > arbitrary @size arguments, which may be much larger than the preferable
> > > I/O size, or may cross extent/device boundaries. Let the lower layers
> > > handle the plugging. Otherwise the plugging code here will turn the
> > > low level plugging into no-ops.
> >
> > I assume you have some numbers to back this up, right? Care to share
> > those?
>
> Yes please.
>
> We've broken this stuff a few times recently - we should review and
> test carefully.

Yes sure. Last time I posted the patch, I did some tests and found no
performance changes. Now for 3.3, the tests started days ago have not
finished now (partly because it is stalled for quite long time due to
unknown reason). The now-available numbers for bs=4k dd's look fine.
The pending tests are for bs=1M dd's and some random fio workloads.

$ ./compare bay/*/*-{3.3.0,3.3.0-plug+}
3.3.0 3.3.0-plug+
------------------------ ------------------------
196.94 -0.0% 196.86 bay/JBOD-2HDD-thresh=1000M/btrfs-10dd-1-3.3.0
196.67 +0.2% 196.99 bay/JBOD-2HDD-thresh=1000M/btrfs-1dd-1-3.3.0
134.34 -0.1% 134.26 bay/JBOD-2HDD-thresh=1000M/ext3-100dd-1-3.3.0
141.75 +0.1% 141.92 bay/JBOD-2HDD-thresh=1000M/ext3-10dd-1-3.3.0
166.74 +0.1% 166.89 bay/JBOD-2HDD-thresh=1000M/ext3-1dd-1-3.3.0
165.90 -0.1% 165.69 bay/JBOD-2HDD-thresh=1000M/ext4-100dd-1-3.3.0
186.83 +0.9% 188.60 bay/JBOD-2HDD-thresh=1000M/ext4-10dd-1-3.3.0
193.30 -0.1% 193.04 bay/JBOD-2HDD-thresh=1000M/ext4-1dd-1-3.3.0
168.34 +0.1% 168.48 bay/JBOD-2HDD-thresh=1000M/xfs-100dd-1-3.3.0
193.43 -0.3% 192.91 bay/JBOD-2HDD-thresh=1000M/xfs-10dd-1-3.3.0
197.59 -0.2% 197.14 bay/JBOD-2HDD-thresh=1000M/xfs-1dd-1-3.3.0
197.09 -0.1% 196.90 bay/JBOD-2HDD-thresh=100M/btrfs-10dd-1-3.3.0
197.20 +0.1% 197.48 bay/JBOD-2HDD-thresh=100M/btrfs-1dd-1-3.3.0
139.07 -0.6% 138.24 bay/JBOD-2HDD-thresh=100M/ext3-10dd-1-3.3.0
157.10 -0.1% 156.99 bay/JBOD-2HDD-thresh=100M/ext3-1dd-1-3.3.0
113.36 -0.6% 112.73 bay/JBOD-2HDD-thresh=100M/ext4-100dd-1-3.3.0
159.88 -0.5% 159.09 bay/JBOD-2HDD-thresh=100M/ext4-10dd-1-3.3.0
191.53 +0.3% 192.09 bay/JBOD-2HDD-thresh=100M/ext4-1dd-1-3.3.0
83.65 +0.4% 83.99 bay/JBOD-2HDD-thresh=100M/xfs-100dd-1-3.3.0
163.12 +0.3% 163.54 bay/JBOD-2HDD-thresh=100M/xfs-10dd-1-3.3.0
194.25 +0.3% 194.88 bay/JBOD-2HDD-thresh=100M/xfs-1dd-1-3.3.0
194.98 +0.2% 195.34 bay/RAID0-2HDD-thresh=1000M/btrfs-10dd-1-3.3.0
194.78 +0.1% 194.96 bay/RAID0-2HDD-thresh=1000M/btrfs-1dd-1-3.3.0
133.07 -0.1% 132.95 bay/RAID0-2HDD-thresh=1000M/ext3-10dd-1-3.3.0
136.90 -0.8% 135.82 bay/RAID0-2HDD-thresh=1000M/ext3-1dd-1-3.3.0
167.61 -1.2% 165.65 bay/RAID0-2HDD-thresh=1000M/ext4-100dd-1-3.3.0
188.04 +1.1% 190.10 bay/RAID0-2HDD-thresh=1000M/ext4-10dd-1-3.3.0
187.74 -0.3% 187.27 bay/RAID0-2HDD-thresh=1000M/ext4-1dd-1-3.3.0
169.01 +0.5% 169.80 bay/RAID0-2HDD-thresh=1000M/xfs-100dd-1-3.3.0
192.03 +0.0% 192.10 bay/RAID0-2HDD-thresh=1000M/xfs-10dd-1-3.3.0
196.06 +0.0% 196.14 bay/RAID0-2HDD-thresh=1000M/xfs-1dd-1-3.3.0
189.04 -0.6% 187.92 bay/RAID0-2HDD-thresh=100M/btrfs-10dd-1-3.3.0
183.23 -0.6% 182.05 bay/RAID0-2HDD-thresh=100M/btrfs-1dd-1-3.3.0
137.19 -0.1% 137.12 bay/RAID0-2HDD-thresh=100M/ext3-10dd-1-3.3.0
140.61 -0.1% 140.54 bay/RAID0-2HDD-thresh=100M/ext3-1dd-1-3.3.0
114.32 -2.3% 111.71 bay/RAID0-2HDD-thresh=100M/ext4-100dd-1-3.3.0
162.42 -0.1% 162.31 bay/RAID0-2HDD-thresh=100M/ext4-10dd-1-3.3.0
183.84 -0.9% 182.19 bay/RAID0-2HDD-thresh=100M/ext4-1dd-1-3.3.0
89.10 -0.3% 88.84 bay/RAID0-2HDD-thresh=100M/xfs-100dd-1-3.3.0
162.48 -0.2% 162.20 bay/RAID0-2HDD-thresh=100M/xfs-10dd-1-3.3.0
188.15 +0.5% 189.12 bay/RAID0-2HDD-thresh=100M/xfs-1dd-1-3.3.0
97.79 -0.0% 97.76 bay/RAID1-2HDD-thresh=1000M/btrfs-10dd-1-3.3.0
97.69 +0.0% 97.71 bay/RAID1-2HDD-thresh=1000M/btrfs-1dd-1-3.3.0
69.84 -1.1% 69.05 bay/RAID1-2HDD-thresh=1000M/ext3-10dd-1-3.3.0
76.62 -0.0% 76.59 bay/RAID1-2HDD-thresh=1000M/ext3-1dd-1-3.3.0
83.97 -0.0% 83.95 bay/RAID1-2HDD-thresh=1000M/ext4-100dd-1-3.3.0
94.87 -0.1% 94.79 bay/RAID1-2HDD-thresh=1000M/ext4-10dd-1-3.3.0
96.87 +0.6% 97.46 bay/RAID1-2HDD-thresh=1000M/ext4-1dd-1-3.3.0
84.57 -0.1% 84.50 bay/RAID1-2HDD-thresh=1000M/xfs-100dd-1-3.3.0
95.51 +0.2% 95.67 bay/RAID1-2HDD-thresh=1000M/xfs-10dd-1-3.3.0
97.44 -0.3% 97.12 bay/RAID1-2HDD-thresh=1000M/xfs-1dd-1-3.3.0
97.69 -0.1% 97.56 bay/RAID1-2HDD-thresh=100M/btrfs-10dd-1-3.3.0
97.75 +0.1% 97.87 bay/RAID1-2HDD-thresh=100M/btrfs-1dd-1-3.3.0
66.68 -0.3% 66.45 bay/RAID1-2HDD-thresh=100M/ext3-10dd-1-3.3.0
76.23 -0.7% 75.73 bay/RAID1-2HDD-thresh=100M/ext3-1dd-1-3.3.0
59.36 +0.9% 59.93 bay/RAID1-2HDD-thresh=100M/ext4-100dd-1-3.3.0
82.63 -0.1% 82.56 bay/RAID1-2HDD-thresh=100M/ext4-10dd-1-3.3.0
95.77 +0.1% 95.86 bay/RAID1-2HDD-thresh=100M/ext4-1dd-1-3.3.0
43.77 -0.1% 43.71 bay/RAID1-2HDD-thresh=100M/xfs-100dd-1-3.3.0
84.24 +0.3% 84.48 bay/RAID1-2HDD-thresh=100M/xfs-10dd-1-3.3.0
97.81 -0.1% 97.76 bay/RAID1-2HDD-thresh=100M/xfs-1dd-1-3.3.0
99.27 -0.0% 99.25 bay/thresh=1000M/btrfs-10dd-1-3.3.0
99.39 +0.0% 99.39 bay/thresh=1000M/btrfs-1dd-1-3.3.0
69.75 +0.2% 69.86 bay/thresh=1000M/ext3-100dd-1-3.3.0
71.79 -0.0% 71.76 bay/thresh=1000M/ext3-10dd-1-3.3.0
80.00 -0.0% 79.99 bay/thresh=1000M/ext3-1dd-1-3.3.0
89.38 -0.1% 89.30 bay/thresh=1000M/ext4-100dd-1-3.3.0
96.43 +0.2% 96.63 bay/thresh=1000M/ext4-10dd-1-3.3.0
98.57 -0.3% 98.28 bay/thresh=1000M/ext4-1dd-1-3.3.0
93.97 -0.2% 93.76 bay/thresh=1000M/xfs-100dd-1-3.3.0
97.51 +0.1% 97.57 bay/thresh=1000M/xfs-10dd-1-3.3.0
99.68 +0.1% 99.74 bay/thresh=1000M/xfs-1dd-1-3.3.0
97.82 -0.2% 97.64 bay/thresh=1000M:990M/btrfs-10dd-1-3.3.0
98.93 +0.7% 99.58 bay/thresh=1000M:990M/btrfs-1dd-1-3.3.0
69.60 -0.0% 69.58 bay/thresh=1000M:990M/ext3-100dd-1-3.3.0
72.97 -0.2% 72.83 bay/thresh=1000M:990M/ext3-10dd-1-3.3.0
79.76 +0.3% 79.99 bay/thresh=1000M:990M/ext3-1dd-1-3.3.0
90.58 -2.1% 88.70 bay/thresh=1000M:990M/ext4-100dd-1-3.3.0
95.27 -0.2% 95.06 bay/thresh=1000M:990M/ext4-10dd-1-3.3.0
96.91 +1.6% 98.48 bay/thresh=1000M:990M/ext4-1dd-1-3.3.0
93.79 -0.4% 93.41 bay/thresh=1000M:990M/xfs-100dd-1-3.3.0
97.07 -0.4% 96.70 bay/thresh=1000M:990M/xfs-10dd-1-3.3.0
99.09 -0.0% 99.04 bay/thresh=1000M:990M/xfs-1dd-1-3.3.0
96.28 -0.2% 96.06 bay/thresh=1000M:999M/btrfs-10dd-1-3.3.0
94.83 +1.4% 96.15 bay/thresh=1000M:999M/btrfs-1dd-1-3.3.0
69.70 +0.2% 69.81 bay/thresh=1000M:999M/ext3-100dd-1-3.3.0
72.79 -0.3% 72.55 bay/thresh=1000M:999M/ext3-10dd-1-3.3.0
79.75 -0.1% 79.69 bay/thresh=1000M:999M/ext3-1dd-1-3.3.0
87.95 +1.6% 89.38 bay/thresh=1000M:999M/ext4-100dd-1-3.3.0
95.18 +0.0% 95.20 bay/thresh=1000M:999M/ext4-10dd-1-3.3.0
98.58 -2.3% 96.36 bay/thresh=1000M:999M/ext4-1dd-1-3.3.0
92.54 -0.1% 92.46 bay/thresh=1000M:999M/xfs-100dd-1-3.3.0
96.81 -0.4% 96.42 bay/thresh=1000M:999M/xfs-10dd-1-3.3.0
99.29 -0.4% 98.91 bay/thresh=1000M:999M/xfs-1dd-1-3.3.0
99.03 +0.1% 99.13 bay/thresh=100M/btrfs-10dd-1-3.3.0
99.55 -0.1% 99.40 bay/thresh=100M/btrfs-1dd-1-3.3.0
55.68 -0.4% 55.48 bay/thresh=100M/ext3-100dd-1-3.3.0
71.40 -0.6% 70.95 bay/thresh=100M/ext3-10dd-1-3.3.0
79.82 +0.1% 79.88 bay/thresh=100M/ext3-1dd-1-3.3.0
63.21 -0.4% 62.99 bay/thresh=100M/ext4-100dd-1-3.3.0
84.42 +0.0% 84.45 bay/thresh=100M/ext4-10dd-1-3.3.0
95.65 +1.9% 97.44 bay/thresh=100M/ext4-1dd-1-3.3.0
49.61 +0.5% 49.87 bay/thresh=100M/xfs-100dd-1-3.3.0
87.76 -0.0% 87.76 bay/thresh=100M/xfs-10dd-1-3.3.0
99.34 +0.0% 99.36 bay/thresh=100M/xfs-1dd-1-3.3.0
92.98 -0.7% 92.29 bay/thresh=10M/btrfs-10dd-1-3.3.0
89.07 +0.4% 89.44 bay/thresh=10M/btrfs-1dd-1-3.3.0
61.76 +0.1% 61.80 bay/thresh=10M/ext3-10dd-1-3.3.0
75.72 +0.3% 75.96 bay/thresh=10M/ext3-1dd-1-3.3.0
67.89 -0.2% 67.77 bay/thresh=10M/ext4-10dd-1-3.3.0
92.00 -0.3% 91.73 bay/thresh=10M/ext4-1dd-1-3.3.0
69.71 -0.4% 69.45 bay/thresh=10M/xfs-10dd-1-3.3.0
92.23 +1.8% 93.89 bay/thresh=10M/xfs-1dd-1-3.3.0
2.60 +7.1% 2.78 bay/thresh=1M/btrfs-10dd-1-3.3.0
3.72 -12.5% 3.25 bay/thresh=1M/btrfs-1dd-1-3.3.0
62.41 +0.3% 62.60 bay/thresh=1M/ext3-10dd-1-3.3.0
58.61 +0.4% 58.82 bay/thresh=1M/ext3-1dd-1-3.3.0
70.70 -1.1% 69.92 bay/thresh=1M/ext4-10dd-1-3.3.0
84.81 +0.1% 84.90 bay/thresh=1M/ext4-1dd-1-3.3.0
66.05 -0.1% 65.97 bay/thresh=1M/xfs-10dd-1-3.3.0
74.81 +0.1% 74.85 bay/thresh=1M/xfs-1dd-1-3.3.0
13467.55 -0.0% 13460.98 TOTAL write_bw

> > > Also moves unplugging for direct I/O from around ->direct_IO() down to
> > > do_blockdev_direct_IO().
> >
> > This could be handled in a separate patch.
>
> I think that would be better.

OK. I'll split it on next post, after the tests complete.

Thanks,
Fengguang
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/