Most recent kernel where this bug did not occur: unknown Distribution: gentoo Hardware Environment: x86_64 Dual Core AMD Opteron Processor 265, LSI20320RB (SN:P055445105) ioc0: 53C1030: Capabilities={Initiator,Target} scsi2 : ioc0: LSI53C1030, FwRev=01032700h, Ports=1, MaxQ=255, IRQ=18 Vendor: IFT Model: A16U-G2421 Rev: 347D Jun 22 08:13:13 db2 kernel: Type: Direct-Access ANSI SCSI revision: 03 Software Environment: gentoo Problem Description: Heavy disk I/O causes MPT driver to produce the following error. ERROR - Chain Buffer Needed ERROR - getFreeChainBuffer Jun 22 09:26:40 db2 kernel: mptscsih: ioc0: SG: Chain Required! sg done 3 Jun 22 09:26:40 db2 kernel: mptscsih: ioc0: ERROR - Chain Buffer Needed, RequestNB=3 sgeOffset=84 Jun 22 09:26:40 db2 kernel: mptscsih: ioc0: getFreeChainBuffer called Jun 22 09:26:40 db2 kernel: mptscsih: ioc0: ERROR - getFreeChainBuffer chainBuf=ffff810103908920 ChainBuffer=ffff8101039087a0 offset=384 chain_idx=4 Jun 22 09:26:40 db2 kernel: Current buff @ ffff810103906754 (index 0xa8)<6> Chain buff @ ffff810103908920 (index 0x4) Jun 22 09:26:40 db2 kernel: mptscsih: ioc0: SG: Chain Required! sg done 7 kern.log with mpt debug will be attached. Steps to reproduce: cp or mv a largeish (>8Gb) file.
Created attachment 8378 [details] extract from kern.log file showing error I have attached an extract. I have the original kern.log if this would be more helpful, but it is very large!
Created attachment 8379 [details] bzip2 of kern.log extract
At the time of raising this bug, I was seeing lots of disk activity and kern.log activity. Having completed the copy of 100 * 8.4Gb files, I am now seeing none, even though I am still copying largish files. Every now and again, I see disk activity, and at the same time, kern.log activity. It is as if the driver is stalling and every now and again, unstalling itself if that makes sense.
this may be related to bug 6739 http://bugzilla.kernel.org/show_bug.cgi?id=6739
Any updates on this bug? Thanks.
GSD, if you still experience this problem with latest kernel, please reopen the bug. Closing since no recent activity.