Alexander Wilms
2007-Jan-03 17:25 UTC
[Xen-devel] Possible Bug in CPU detection / clock skew?
Hi all, I have a very strange clock skew problem on my Server since running Xen 3.0.3 that I have never seen before on the same machine running Xen 2.0.x. Now I compared boot messages from the old and new installation and I saw a huge difference in the reported CPU speeds under 3.0.3 hypervisor. Could this be a reason for the clock skew? Excuse me if the question is stupid, I''m not a CPU expert... Xen 2.0.x: Boot1: <6>Xen reported: 498.762 MHz processor Boot2: <6>Xen reported: 498.765 MHz processor Xen 3.0.3: Boot1: <6>Xen reported: 556.714 MHz processor Boot2: <6>Xen reported: 621.840 MHz processor Hardware is a Dual CPU Intel 440BX Server Board with 2 PIII 500 Cpu''s (I know, it''s veery oooold...) Thanks, Alex _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Keir Fraser
2007-Jan-04 10:10 UTC
Re: [Xen-devel] Possible Bug in CPU detection / clock skew?
On 3/1/07 17:25, "Alexander Wilms" <alex.wilms@adminguru.org> wrote:> I have a very strange clock skew problem on my Server since running Xen 3.0.3 > that I have never seen before on the same machine running Xen 2.0.x. > > Now I compared boot messages from the old and new installation and I saw a > huge difference in the reported CPU speeds under 3.0.3 hypervisor. Could this > be a reason for the clock skew? Excuse me if the question is stupid, I''m not > a CPU expert...It does seem likely. The old messages were correct and the new ones are wrong, for some reason. I don;t think the calibration code has changed significantly but I''d have to check that. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Hello all, seems that I found another bug regarding tx checksum. Had a problem with DNAT''ing ftp. Found out that the ip_nat_ftp module didn''t rewrite the pasv source address correctly. With wireshark I could see that this happened because of excessive TCP retransmissions due to wrong checksum. My Xen/network setup looks like this: ftpserver --- firewall ---internet While the ftp server runs in dom0. Dom0 is bridged. Firewall is a domU with delegated nics. Xen is the version released with openSUSE 10.2: xm info host : xen0 release : 2.6.18.2-34-xen version : #1 SMP Mon Nov 27 11:46:27 UTC 2006 machine : i686 nr_cpus : 2 nr_nodes : 1 sockets_per_node : 2 cores_per_socket : 1 threads_per_core : 1 cpu_mhz : 498 hw_caps : 0383fbff:00000000:00000000:00000040 total_memory : 1023 free_memory : 1 max_free_memory : 139 max_para_memory : 135 max_hvm_memory : 126 xen_major : 3 xen_minor : 0 xen_extra : .3_11774-20 xen_caps : xen-3.0-x86_32 xen_pagesize : 4096 platform_params : virt_start=0xfc000000 xen_changeset : 11774 cc_compiler : gcc version 4.1.2 20061115 (prerelease) (SUSE Linux) cc_compile_by : abuild cc_compile_domain : suse.de cc_compile_date : Tue Nov 28 09:28:29 UTC 2006 xend_config_format : 2 Workaround/Solution: Execute ethtool -K tx off eth0 in dom0 and the problem disappears. If needed network dumps are available upon request. Alex _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel