On 02/22/2012 09:59 PM, Eric Blake wrote:
Output is still in kibibytes, but input can now be in different
scales for ease of typing.
* src/conf/domain_conf.c (virDomainParseMemory): New helper.
(virDomainDefParseXML): Use it when parsing.
* docs/schemas/domaincommon.rng: Expand XML.
* docs/formatdomain.html.in (elementsMemoryAllocation): Document
scaling.
---
docs/formatdomain.html.in | 39 ++++++++++---
docs/schemas/domaincommon.rng | 4 +-
src/conf/domain_conf.c | 129 ++++++++++++++++++++++++++++++++++------
3 files changed, 142 insertions(+), 30 deletions(-)
self-NACK to v1 - I forgot to add a unit test that exercises the new
unit parsing, and...
+ tebibytes (1,099,511,627,776). However, the value will be
+ rounded up to the nearest kibibyte by libvirt, and may be
+ further rounded to the granularity supported by the
+ hypervisor. As a sanity check, values less than 4000KiB are
+ not permitted.
This statement is not true (I originally tried to make the C code match
the RNG, but it caused too many testsuite failures, so I rebased and
inserted 2/3 to fix the RNG instead, but forgot to touch this up again).
I'll fix that in v2, along with a rename to unit='KiB' instead of units.
--
Eric Blake eblake(a)redhat.com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org