>>>>> "Roger" == Roger Willcocks <roger@xxxxxxxxxxxxxxxx> writes: Raghava, This appears to be a regression after the latest update... Roger> I've see this on multiple machines here with various kernels and Roger> hardware configurations, the most canonical being: Roger> hardware: HP Z840 + ASR8405 (4 x HGST raid 6 formatted as xfs) Roger> Z840 firmware: v01.16 ASR8405 firmware: 7.8-0 (32730) Roger> kernel: 4.5.4.rc4 tip with command line options 'ro single' Roger> (Adaptec driver 1.2-1 (41024)) Roger> However the same thing happens with 2 and 4 x ASR81605ZQ on a Roger> Supermicro motherboard with the Adaptec 1.2.1-40300 driver. The Roger> 30200 driver, fwiw, causes a kernel panic (page fault in Roger> aac_intr_normal). Roger> To reproduce, queue up a bunch of write requests: Roger> # dd if=/dev/zero of=/file/on/raid & Roger> then in a loop, run 'arcconf getversion' repeatedly. I'm using Roger> arcconf 1.7 (B21229) Roger> Within thirty seconds or so (less than 100 repeats) arcconf hangs Roger> with userspace stack: Roger> #0 0x00007f60103a9a47 in ioctl () from /lib64/libc.so.6 #1 Roger> 0x000000000059f4bc in faos_SendReceiveFIB(FSAAPI_CONTEXT*, _FIB*, Roger> unsigned int) () #2 0x000000000056de3f in Roger> FsaInternalSendReceiveFib(FSAAPI_CONTEXT*, _FIB*, int, unsigned Roger> int) () #3 0x000000000058da76 in FsaGetAdapterSasPhyInfo () #4 Roger> 0x00000000004f97ea in ArcAdapter::initSasPhyInfo(Ret&) () #5 Roger> 0x0000000000507516 in ArcAdapter::buildChildren(Ret&) () #6 Roger> 0x0000000000533486 in ArcSystem::buildChildren(StorLib*, Ret&) () Roger> #7 0x0000000000492170 in StorLib::getSystemConfig() () #8 Roger> 0x00000000004289c0 in main () Roger> The kernel's stuck in down_interruptible() called from Roger> aac_fib_send(). ^C releases it (but you can repeat the Roger> experiment.) Roger> I've done a small amount of digging and afaict the controller Roger> issues an interrupt when the request completes, but the driver Roger> doesn't call up() because the XFerState in the reply doesn't Roger> match to the request. Roger> sent: 1010a1 rcvd: 804832ad Roger> I don't really know enough about the controller protocol to debug Roger> this further. Roger> First things first I'd be grateful if somebody could confirm the Roger> hang. -- Martin K. Petersen Oracle Linux Engineering -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
