[Bug 1899621] Re: [Ubuntu 20.04] zlib: inflateSyncPoint() returns an incorrect result on z15
Łukasz Zemczak
1899621 at bugs.launchpad.net
Thu Oct 29 23:22:37 UTC 2020
I see that the burp armhf autopkgtest keeps on failing on the new zlib.
I wonder if the burp-unit-test suite is really flaky and the release-
pocket run only passed by chance?
--
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to zlib in Ubuntu.
https://bugs.launchpad.net/bugs/1899621
Title:
[Ubuntu 20.04] zlib: inflateSyncPoint() returns an incorrect result on
z15
Status in Ubuntu on IBM z Systems:
Fix Committed
Status in zlib package in Ubuntu:
Fix Released
Status in zlib source package in Focal:
Fix Committed
Status in zlib source package in Groovy:
Fix Released
Bug description:
Description: zlib: inflateSyncPoint() returns an incorrect result on z15
Symptom: Certain rsync builds fail with "error in rsync protocol data
stream" on z15.
Problem: inflateSyncPoint() does not take into account the hardware
compression state and returns an incorrect result.
Solution: Make inflateSyncPoint() fail if the hardware compression is on.
The hardware does not provide enough information in order to
implement this function.
[Impact]
* Certain rsync builds fail with "error in rsync protocol data
stream" on z15.
* On ubuntu, rsync normally uses zstd or lz4. But when rsync is
forced to use non-default zlib compression (-z flag) it uses
inflateSyncPoint() API. This can also happen when rsync on the the
other end doesn't support zstd & lz4.
* inflateSyncPoint() does not take into account the hardware
compression state and returns an incorrect result.
* Make inflateSyncPoint() fail if the hardware compression is on. The
hardware does not provide enough information in order to implement
this function.
* Above makes rsync to succeed, when zlib uses hardware compression.
[Test Case]
Reproduction: z15 only: printf 1 >1 && rsync -z 1 2
[Regression Potential]
* inflateSyncPoint API is rarely used. Scanning codesearch, there is
a lot of false positives due to embedded copies of zlib in all the
things. I do see that both rsync and backuppc-rsync use it. It used
during a safety check to ensure that algorithm is not at a sync point
(or that cannot be determined). Nonetheless, returning error is a
safer implementation for this API call.
[Other Info]
* This is a regression introduced by adding & enabling zlib hw
acceleration by default on z15; and discovering using rsync that one
API is implemented incorrectly.
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu-z-systems/+bug/1899621/+subscriptions
More information about the foundations-bugs
mailing list