#10850 NORM 12.1.0: zhashfs: use FIEMAP for detecting unused disk blocks
Zarro Boogs per Child
bugtracker at laptop.org
Thu Mar 8 11:38:31 EST 2012
#10850: zhashfs: use FIEMAP for detecting unused disk blocks
------------------------------------+---------------------------------------
Reporter: dsd | Owner: dsd
Type: enhancement | Status: closed
Priority: normal | Milestone: 12.1.0
Component: build-system | Version: not specified
Resolution: fixed | Keywords:
Next_action: code | Verified: 0
Deployment_affected: | Blockedby:
Blocking: |
------------------------------------+---------------------------------------
Changes (by dsd):
* status: new => closed
* resolution: => fixed
* milestone: Future Release => 12.1.0
Comment:
On new F17 builds, the resultant disk image includes holes - something
(mke2fs?) is actively punching holes in the disk image, losing the marker
that was put in place. So I've implemented the FIEMAP based approach which
avoids this issue.
I also tested that it performs the same as the existing format, by
backporting it to olpc-os-builder 4.0.0 and rebuilding 11.3.0 four times,
twice with the previous fill-based system and twice with FIEMAP.
Values are: filesize of .zd image, line count of .zsp file (i.e. number of
blocks)
{{{
Test 1 with existing fill-pattern code:
2gb: 761274739, 12885
4gb: 762773004, 13377
8gb: 764029458, 14365
Test 2 with existing fill-pattern code:
2gb: 761222412, 12887
4gb: 762941008, 13379
8gb: 763959516, 14366
Test 1 with FIEMAP:
2gb: 760980361, 12881
4gb: 762623007, 13371
8gb: 763506946, 14360
Test 2 with FIEMAP:
2gb: 761056419, 12880
4gb: 762560031, 13370
8gb: 763454544, 14361
}}}
Overall the FIEMAP approach does not regress over the fill-pattern method;
it could even be argued that FIEMAP is ever so slightly better.
--
Ticket URL: <http://dev.laptop.org/ticket/10850#comment:2>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system
More information about the Bugs
mailing list