> Simon - the RH9 kernels, where we turned O_DIRECT back on -only- for
> xfs, together with the nptl patches in the RH9 kernel, seems to cause
> problems with RPM - even if the RPM db is not on xfs.
Hmm, and this happens on all x86 platforms?
I'm the maintainer of cyrus-imapd rpms and there was already a big
surprise with nptl kernels in combination with db4. Interestingly the
problem appeared only on AMD K6 and Athlon CPU's. The following error from
cyrus-imapd showed up in syslog:
May 15 13:51:25 co ctl_cyrusdb: DBERROR: dbenv->open
'/var/lib/imap/db' failed: Function not implemented
This was with the original RedHat kernels as well as on XFS enabled ones.
The solution was to rebuild db4 without --enable-posixmutexes. A modified
spec to rebuild db4 from RedHat is here
Is it possible that the nonnptl db4 package is a solution for the problem
discussed here? I'll just try it but can someone tell me what problem I
should expect if this fix doesn't work. Does rpm not work at all?
> FWIW, the other solution seems to be to get the latest RPMs from rpm.org
> (4.2.x) which turn off O_DIRECT in the rpm source.
> I still don't know what the underlying problem actually is.
> On Tue, 2003-08-26 at 09:51, Simon Matter wrote:
>> Hmm, can somebody explain what the patch 1300 problem is? I have built
>> my own 2.4.20-20.7.XFS1.3.0 and 2.4.20-20.9.XFS1.3.0 rpms and I'm using
>> the 20.7 version on RedHat 7.2 without any problem. Did I miss
>> something? Is there a problem when using in on RedHat 9? Of course 20.9
>> has the nptl patches so should I expect any problems which are not
>> present in 20.7?
> Eric Sandeen [C]XFS for Linux http://oss.sgi.com/projects/xfs
> sandeen@xxxxxxx SGI, Inc. 651-683-3102
Simon Matter Tel: +41 61 695 57 35
Fr.Sauter AG / CIT Fax: +41 61 695 53 30
Im Surinam 55
CH-4016 Basel [mailto:simon.matter@xxxxxxxxxxxxxxxx]