On Mon, May 31, 2021 at 10:36:11AM +0200, Michal Privoznik wrote:
This test was generated on a guest with the following NUMA
configuration:
<numa>
<cell id='0' cpus='0-23' memory='4194304'
unit='KiB' discard='yes'>
<cache level='1' associativity='direct'
policy='writeback'>
<size value='10' unit='KiB'/>
<line value='8' unit='B'/>
</cache>
<cache level='2' associativity='full'
policy='writethrough'>
<size value='128' unit='KiB'/>
<line value='16' unit='B'/>
</cache>
</cell>
<cell id='1' memory='2097152' unit='KiB'>
<cache level='1' associativity='direct'
policy='writeback'>
<size value='10' unit='KiB'/>
<line value='8' unit='B'/>
</cache>
</cell>
<interconnects>
<latency initiator='0' target='0' type='access'
value='5'/>
<latency initiator='0' target='0' type='read'
value='6'/>
<latency initiator='0' target='0' type='write'
value='7'/>
<latency initiator='0' target='1' type='access'
value='10'/>
<latency initiator='0' target='1' type='read'
value='11'/>
<latency initiator='0' target='1' type='write'
value='12'/>
<bandwidth initiator='0' target='0' type='access'
value='204800' unit='KiB'/>
<bandwidth initiator='0' target='0' type='read'
value='205824' unit='KiB'/>
<bandwidth initiator='0' target='0' type='write'
value='206848' unit='KiB'/>
<bandwidth initiator='0' target='0' cache='1'
type='access' value='208896' unit='KiB'/>
<bandwidth initiator='0' target='0' cache='1'
type='read' value='209920' unit='KiB'/>
<bandwidth initiator='0' target='0' cache='1'
type='write' value='210944' unit='KiB'/>
<bandwidth initiator='0' target='1' type='access'
value='102400' unit='KiB'/>
<bandwidth initiator='0' target='1' type='read'
value='103424' unit='KiB'/>
<bandwidth initiator='0' target='1' type='write'
value='104448' unit='KiB'/>
<bandwidth initiator='0' target='1' cache='1'
type='access' value='105472' unit='KiB'/>
<bandwidth initiator='0' target='1' cache='1'
type='read' value='106496' unit='KiB'/>
<bandwidth initiator='0' target='1' cache='1'
type='write' value='107520' unit='KiB'/>
</interconnects>
</numa>
The sysfs content was also copied over from the VM but only those
files which are accessed in the test are stored in the repo.
Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com>
---
This unfortunately introduces symbolic link loops. They are fine in
sysfs, but we always had issues with them when running distcheck and
similar.
I tried it now and did not reproduce the exact issue, but it still is
not nice (global -u complains, some programs might just end up in an
infinite loop). Historically we replicated "just enough" to use it in
tests, but not so much that it would introduce loops. I would prefer if
it stayed that way.