The LSI SAS 9201 (chip: LSI SAS2116) device in my system isn't detecting any connected disks. Relevant boot dmesg, with 7 disks attached: Fusion MPT base driver 3.04.20 Copyright (c) 1999-2008 LSI Corporation Fusion MPT SAS Host driver 3.04.20 mpt2sas version 16.100.00.00 loaded scsi9 : Fusion MPT SAS Host mpt2sas0: 64 BIT PCI BUS DMA ADDRESSING SUPPORTED, total mem (24692096 kB) mpt2sas0: IO-APIC enabled: IRQ 16 mpt2sas0: iomem(0x00000000fbbfc000), mapped(0xffffc90008380000), size(16384) mpt2sas0: ioport(0x000000000000de00), size(256) mpt2sas0: Allocated physical memory: size(13309 kB) mpt2sas0: Current Controller Queue Depth(5333), Max Controller Queue Depth(10392) mpt2sas0: Scatter Gather Elements per IO(128) mpt2sas0: LSISAS2116: FWVersion(09.00.00.00), ChipRevision(0x02), BiosVersion(00.00.00.00) mpt2sas0: Protocol=(Initiator,Target), Capabilities=(TLR,EEDP,Snapshot Buffer,Diag Trace Buffer,Task Set Full,NCQ) mpt2sas0: sending port enable !! mpt2sas0: host_add: handle(0x0001), sas_addr(0x5000d31000407c1f), phys(16) mpt2sas0: port enable: SUCCESS This was as an auto-loaded module. I can remove it, waiting module removal not supported: please upgrade mpt2sas version 16.100.00.00 unloading mpt2sas0: _base_make_ioc_ready mpt2sas0: _base_make_ioc_ready: ioc_state(0x24000000) mpt2sas0: sending message unit reset !! mpt2sas0: _base_wait_for_doorbell_ack: successful count(1), timeout(15) mpt2sas0: message unit reset: SUCCESS And reload with higher logging level (Enabled "firmware events and reply with additional info", "handshake and initialization", "application using IOCTLS", "host reset and task management" logging). Note that this reinsert is with two ports connected to seven disks total: setting logging_level(0x0000a738) mpt2sas version 16.100.00.00 loaded scsi11 : Fusion MPT SAS Host mpt2sas0: mpt2sas_base_attach mpt2sas0: mpt2sas_base_map_resources mpt2sas0: 64 BIT PCI BUS DMA ADDRESSING SUPPORTED, total mem (24692096 kB) mpt2sas0: msix is supported, vector_count(1) mpt2sas0: IO-APIC enabled: IRQ 16 mpt2sas0: iomem(0x00000000fbbfc000), mapped(0xffffc900077d8000), size(16384) mpt2sas0: ioport(0x000000000000de00), size(256) mpt2sas0: _base_get_ioc_facts mpt2sas0: _base_wait_for_doorbell_int: successful count(1), timeout(5) mpt2sas0: _base_wait_for_doorbell_ack: successful count(1), timeout(5) (prev line 3 more times) mpt2sas0: _base_wait_for_doorbell_int: successful count(0), timeout(5) mpt2sas0: _base_wait_for_doorbell_int: successful count(1), timeout(5) (prev line 31 more times) mpt2sas0: _base_wait_for_doorbell_not_used: successful count(0), timeout(5) offset:data [0x00]:03100200 [0x04]:00001200 [0x08]:00000000 [0x0c]:00000000 [0x10]:00000000 [0x14]:00010480 [0x18]:22132898 [0x1c]:0001285c [0x20]:09000000 [0x24]:00000020 [0x28]:03480004 [0x2c]:00440044 [0x30]:09df0003 [0x34]:0020fff0 [0x38]:008003a0 [0x3c]:00000011 mpt2sas0: hba queue depth(10392), max chains per io(128) mpt2sas0: request frame size(128), reply frame size(128) mpt2sas0: _base_make_ioc_ready mpt2sas0: _base_make_ioc_ready: ioc_state(0x14000000) mpt2sas0: _base_get_port_facts mpt2sas0: _base_wait_for_doorbell_int: successful count(1), timeout(5) (prev line 4 more times) mpt2sas0: _base_wait_for_doorbell_int: successful count(0), timeout(5) mpt2sas0: _base_wait_for_doorbell_int: successful count(1), timeout(5) (prev line 13 more times) mpt2sas0: _base_wait_for_doorbell_not_used: successful count(0), timeout(5) offset:data [0x00]:05070000 [0x04]:00000000 [0x08]:00000000 [0x0c]:00000000 [0x10]:00000000 [0x14]:00003000 [0x18]:0000020e mpt2sas0: _base_allocate_memory_pools mpt2sas0: scatter gather: sge_in_main_msg(1), sge_per_chain(9), sge_per_io(128), chains_per_io(15) mpt2sas0: scsi host: can_queue depth (5333) mpt2sas0: request pool(0xffff8800d8600000): depth(10392), frame_size(128), pool_size(1299 kB) mpt2sas0: request pool: dma(0xd8600000) mpt2sas0: scsiio(0xffff8800d8600000): depth(5333) mpt2sas0: chain pool depth(79995), frame_size(128), pool_size(9999 kB) mpt2sas0: hi_priority(0xffff8800d86a6b00): depth(2527), start smid(5334) mpt2sas0: internal(0xffff8800d86f5a80): depth(2532), start smid(7861) mpt2sas0: sense pool(0xffff8805fed80000): depth(5333), element_size(96), pool_size(499 kB) mpt2sas0: sense_dma(0x5fed80000) mpt2sas0: reply pool(0xffff8805ff400000): depth(10456), frame_size(128), pool_size(1307 kB) mpt2sas0: reply_dma(0x5ff400000) mpt2sas0: reply_free pool(0xffff8800da860000): depth(10456), element_size(4), pool_size(40 kB) mpt2sas0: reply_free_dma(0xda860000) mpt2sas0: reply post free pool(0xffff8800dafc0000): depth(20864), element_size(8), pool_size(163 kB) mpt2sas0: reply_post_free_dma = (0xdafc0000) mpt2sas0: config page(0xffff8800d840a000): size(512) mpt2sas0: config_page_dma(0xd840a000) mpt2sas0: Allocated physical memory: size(13309 kB) mpt2sas0: Current Controller Queue Depth(5333), Max Controller Queue Depth(10392) mpt2sas0: Scatter Gather Elements per IO(128) mpt2sas0: _base_make_ioc_operational mpt2sas0: _base_send_ioc_init offset:data [0x00]:02000004 [0x04]:00000000 [0x08]:00000000 [0x0c]:1c000200 [0x10]:00000000 [0x14]:00000000 [0x18]:00200000 [0x1c]:28d85180 [0x20]:00000005 [0x24]:00000005 [0x28]:d8600000 [0x2c]:00000000 [0x30]:dafc0000 [0x34]:00000000 [0x38]:da860000 [0x3c]:00000000 [0x40]:b5094698 [0x44]:00000144 mpt2sas0: _base_wait_for_doorbell_int: successful count(1), timeout(5) (prev line 20 more times) mpt2sas0: _base_wait_for_doorbell_int: successful count(0), timeout(10) mpt2sas0: _base_wait_for_doorbell_int: successful count(1), timeout(5) (prev line 9 more times) mpt2sas0: _base_wait_for_doorbell_not_used: successful count(0), timeout(5) offset:data [0x00]:02050004 [0x04]:00000000 [0x08]:00000000 [0x0c]:00000000 [0x10]:00000000 mpt2sas0: _base_event_notification mpt2sas0: _base_event_notification: complete mpt2sas0: LSISAS2116: FWVersion(09.00.00.00), ChipRevision(0x02), BiosVersion(00.00.00.00) mpt2sas0: Protocol=(Initiator,Target), Capabilities=(TLR,EEDP,Snapshot Buffer,Diag Trace Buffer,Task Set Full,NCQ) mpt2sas0: sending port enable !! mpt2sas0: Discovery: (start) mpt2sas0: SAS Enclosure Device Status Change mpt2sas0: SAS Topology Change List mpt2sas0: Discovery: (stop) mpt2sas0: discovery event: (start) mpt2sas0: Discovery: (start) mpt2sas0: Discovery: (stop) mpt2sas0: Discovery: (start) mpt2sas0: Discovery: (stop) mpt2sas0: Discovery: (start) mpt2sas0: Discovery: (stop) mpt2sas0: host_add: handle(0x0001), sas_addr(0x5000d31000407c1f), phys(16) mpt2sas0: enclosure status change: (enclosure add) handle(0x0001), enclosure logical id(0x5000d31000407c1f) number slots(0) mpt2sas0: sas topology change: (responding) handle(0x0000), enclosure_handle(0x0001) start_phy(00), count(16) mpt2sas0: updating handles for sas_host(0x5000d31000407c1f) mpt2sas0: discovery event: (stop) mpt2sas0: discovery event: (start) mpt2sas0: discovery event: (stop) mpt2sas0: discovery event: (start) mpt2sas0: discovery event: (stop) mpt2sas0: discovery event: (start) mpt2sas0: discovery event: (stop) mpt2sas0: port enable: complete from worker thread mpt2sas0: port enable: SUCCESS When I then unplug one of the cables, I get no change in dmesg. It doesn't detect any disconnect whatsoever. When I plug the cable back in, likewise -- nothing logged in dmesg. This is a replacement from LSI 3442 card (mptsas driver -- different), so I know the cabling and disks all work perfectly. I upgraded for 4TB disk support, but still testing so haven't put those in play yet. LSI has released an updated driver, 18.00.00.00, as opposed to the 3.13.6 version, 16.100.00.00. However, I lack the ability to convert this driver to the newer kernel. LSI driver, 18.00.00.00: http://www.lsi.com/downloads/Public/Host%20Bus%20Adapters/Host%20Bus%20Adapters%20Common%20Files/SAS_SATA_6G_P18/Linux%20Driver-RH5_SLES10_P18.zip (Caution: 866MB) mpt2sas-18.00.00.00-src.tar.gz (259K), from archive: http://www.2shared.com/file/oYzP_HeF/mpt2sas-18000000-srctar.html
The issue turned out to be hardware: Some combination of flashing the firmware and BIOS caused all disks to register. I did get the 18.00.00.00 driver to compile under 3.13.6, however it gave me its own problems: as I attempted to dd if="$f" of="$f" bs=65536 across every disk in the system simultaneously, the system would, after 20-30 minutes, lock solid. (This process proceeded fine if of=/dev/null; I'm doing this to find disks that I know have sector errors.) I have no dmesg output. Further, modprobe -r was _very_ slow on this driver, eventually printing to dmesg a trace including "warn_slowpath_fmt". Attached is a basic patch to get the new LSI driver to compile under 3.13.6, but no guarantees as to the quality of anything.
Created attachment 130291 [details] Patch to LSI 18.00.00.00 driver to compile on 3.13.6