[OpenAFS-devel] OpenAFS 1.8.7 available

Ken Aaker kenaaker@gmail.com
Thu, 21 Jan 2021 14:34:57 -0600


>From main ignorance, I stumbled about for a week or so. But, I've
finally got my cell running again. I managed that by finding and
installing the 1.8.7 rpm build off the OpenSuSE build servers, and
fumbling my way through the re-keying process.  Now, I've got 4 Debian
10 (buster) VMs that need the same update, and I can't find it?  Any
leads? The latest version I have been able to find at Debian is 1.8.6.5?
2 of the VMs are "Fileservers" and 2 are build systems.

Regards,

Ken Aaker

On 1/14/21 5:40 PM, Benjamin Kaduk wrote:
> The OpenAFS Guardians are happy to announce the availability of OpenAFS 1.8.7.
> Source files can be accessed via the web at:
>
>        https://www.openafs.org/release/openafs-1.8.7.html
>
> or via AFS at:
>
>        UNIX: /afs/grand.central.org/software/openafs/1.8.7/
>        UNC: \\afs\grand.central.org\software\openafs\1.8.7\
>
> This release fixes a critical issue with the generation of Rx connection IDs
> (CIDs) for Rx clients started after 14 Jan 2021 08:25:36 AM UTC (Unix epoch
> time 0x60000000).  Unpatched systems will always use the fixed value of
> 0x80000002 as the CID, which causes connections to fail for multiple reasons.
> Client commands such as `vos examine` will time out if run after that
> time, cache managers started after that time will be unable to access files,
> fileservers started after that time will be unable to access the Ubik databases,
> and database servers started after that time will be unable to participate
> in a quorum or Ubik elections.  In particular, fileservers that restart,
> including due to a scheduled weekly restart, will not be able to register
> with the volume location service or verify group membership with the
> protection service.
>
> The fix causes the initial CID to be randomly generated (without dependence
> on the current time) and removes the faulty logic intended to detect signed
> integer overflow (which is not needed since the field in question is now
> an unsigned integer).
>
> Thanks to Jeffrey Altman of Auristor Inc. for tracking down the key issue.
>
> Bug reports should be filed to openafs-bugs@openafs.org.
>
> Benjamin Kaduk
> for the OpenAFS Guardians
>