Date: Fri, 04 Jun 2004 14:24:01 -0500 From: "J.D. Bronson" <jbronson@wixb.com> To: freebsd-scsi@freebsd.org Subject: Re: mpt driver patch Message-ID: <6.1.1.1.2.20040604142103.00ab1950@localhost> In-Reply-To: <OFFD492855.209A0133-ONC1256EA9.0060A9ED-C1256EA9.00691FCE@ env.cz> References: <OFFD492855.209A0133-ONC1256EA9.0060A9ED-C1256EA9.00691FCE@env.cz>
next in thread | previous in thread | raw e-mail | index | archive | help
At 02:07 PM 06/04/2004, Jan_Rybensky@env.cz wrote: >Hi All, > >as well as the others I've encoutered serious performace problems with IM >volumes on LSI 1030 controller. It's because the mpt driver is unable to >detect hidden disks of the IM volumes and initialize them. That disks >operates at default SCSI speed and slows the whole volume. >This patch is far from perfect (not performs full DV), but it solves the >performance issues. It seems to be stable, but it's tested only on IBM >version of controller (on-board controllers of xSeries 335, 345). > >There is other issue I'm unable to solve: although all discs operating at >proper speeds, the resync of mirrored disks is still very slow. The >controller probably needs additional initializations, but I'm unable to >trace them in linux driver. Can anybody help? Any suggestions are >welcomed. > >Here is diff against 4.9-REL mpt sources (it works well on 4.10-REL too) >and a tarball of the whole mpt directory from /sys/dev with patched driver >sources is here: http://mpt-freebsd.env.cz. > >Many thanks for any feedback. > > Jan Rybensky I have seen this exact same behavior under 5.2.1 ...I have the LSI 20320R card installed. Using normal setup - this card works well and the speeds are as expected: da0 at mpt0 bus 0 target 0 lun 0 da0: <SEAGATE ST336607LW 0007> Fixed Direct Access SCSI-3 device da0: 320.000MB/s transfers (160.000MHz, offset 63, 16bit), Tagged Queueing Enabled da0: 35003MB (71687372 512 byte sectors: 255H 63S/T 4462C) however, like you pointed out - the mirroring is atrocious. It took over 5hrs to fully sync a mirror (RAID-1). Then when I rebooted, the performance was less than 2MB/sec on any transfers to/from the machine. As long as the mirror is not active, the card+drives seem to work just fine. FWIW, setting up a mirror on this same card+drives under solaris x86, works perfect and resync's in just over 2hrs. -- J.D. Bronson Aurora Health Care // Information Services // Milwaukee, WI USA Office: 414.978.8282 // Email: jd@aurora.org // Pager: 414.314.8282
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?6.1.1.1.2.20040604142103.00ab1950>