On Tue, Jul 03, 2018 at 20:41:54 +0800, dujiancheng wrote:
From: Dujiancheng <dujiancheng10(a)gmail.com>
dujiancheng (1):
qemu: Added support L2 table cache for qcow2 disk.
L2 table cache can be set by the new element diskCache.
Use the following methods to set the L2 table cache
and cache clean interval:
<diskCache>
<cache level='2'>
<size unit='KiB'>10240</size>
</cache>
<clean interval='900'/>
</diskCache>
Since you insist on sending your own version rather than reusing the
better previously posted version, here are a few things this
patch is missing:
- capability which allows to check whether qemu supports the new
features
- cache is a property of the image and not the disk, also it needs to be
configurable for <backingStore> so that it's usable for the upcoming
blockdev work
- the documentation does not explain how it's supposed to be used.
- the RNG schema indentation is wrong
- the formatter for the XML should not have any logic .. e.g.
determining that only L2 cache is printed
- the code itself is rather strange and does not allow multiple cache
levels although the data structure suggests so
- the name for the tests are wrong since this is not disk cache but
image cache
- there are no checks whether a format supporting the l2 cache is used
(qcow2 only)
- make syntax-check fails:
--- tests/qemuxml2argvdata/disk-cache.args 2018-07-03 15:29:18.386122485 +0200
+++ - 2018-07-03 15:30:36.263098059 +0200
@@ -22,7 +22,8 @@
-no-acpi \
-boot c \
-usb \
--drive
file=/tmp/data.img,format=raw,l2-cache-size=10240,cache-clean-interval=900,if=none,id=drive-virtio-disk0
\
+-drive file=/tmp/data.img,format=raw,l2-cache-size=10240,\
+cache-clean-interval=900,if=none,id=drive-virtio-disk0 \
-device virtio-blk-pci,bus=pci.0,addr=0x3,drive=drive-virtio-disk0,\
id=virtio-disk0 \
-drive file=/tmp/logs.img,format=raw,if=none,id=drive-virtio-disk1 \
Incorrect line wrapping in tests/qemuxml2argvdata/disk-cache.args
Use test-wrap-argv.pl to wrap test data files
Signed-off-by: dujiancheng <dujiancheng10(a)gmail.com>
I think it was pointed out that we prefer real names...