I reproduced this only with "queue_if_no_path" feature. Ranaldo already showed some places are dusty. A new patchset is due soon. Alasdair, I guess there still are some unpined references to sd_mod on "map-load error-code-path". I did experience refcount going crazy when trying to create a map with path with wrong size (dummy 1GB). regards, cvaroqui Le vendredi 24 d?embre 2004 �6:52 +0100, Hergen Lange a �it : > Hello, > > i try to setup a multipath to a EMC Clariion Device with the following > infrstructure. > Kernel 2.6.10-rc3 and udm2 Patch. > multipath-tools 0.4.1 > > I have one qla2310 Adapter "Firmware version 3.03.02 IPX, Driver version > 8.00.00b21-k" > wich is zoned to SPA Port0 and SPB Port0 of the Clariion. So we have > two path to the LUN. > > I setup multipath.conf with the following. > > multipath { > wwid 3600601601b121200540b8c1caf4ad911 > alias yellow > path_grouping_policy failover > path_checker emc_clariion > path_selector "round-robin 0" > getuid_callout "/sbin/scsi_id -g -u -s %n" > } > device { > vendor "DGC " > product "RAID 5 " > path_grouping_policy failover > getuid_callout "/sbin/scsi_id -g -u -s /block/%n" > path_checker emc_clariion > } > > When i start the multipathd i get the following table > > yellow: 0 20971520 multipath 0 0 2 1 round-robin 0 1 1 8:64 1000 > round-robin 0 1 1 8:80 1000 > > multipath -v2 -l reports > > yellow (3600601601b121200540b8c1caf4ad911) > [size=10 GB][features="0"][hwhandler="0"] > \_ round-robin 0 [enabled][first] > \_ 2:0:1:0 sde 8:64 [ready ][active] > \_ round-robin 0 [enabled] > \_ 2:0:2:0 sdf 8:80 [ready ][active] > > I can work with the device dd if=/dev/yellow of=/dev/null count=1000 > > When i disable one port on the switch to the Clarion, or unplug the cable from > the SP (Clariion) the dd command goes to D+ state (ps ax) > and I get messages like this. > > Dec 23 14:35:31 lxssantm1 kernel: SCSI error : <2 0 1 0> return code = > 0x20000 > Dec 23 14:35:31 lxssantm1 kernel: end_request: I/O error, dev sde, sector 0 > Dec 23 14:35:31 lxssantm1 kernel: end_request: I/O error, dev sde, sector 8 > Dec 23 14:35:31 lxssantm1 multipathd: devmap event on yellow > Dec 23 14:35:45 lxssantm1 multipathd: 8:64 : emc_clariion_checker: query > command indicates error > Dec 23 14:35:45 lxssantm1 kernel: SCSI error : <2 0 1 0> return code = > 0x10000 > Dec 23 14:35:45 lxssantm1 last message repeated 6 times > Dec 23 14:35:45 lxssantm1 kernel: device-mapper: multipath_message > received. argc: 2 > Dec 23 14:35:45 lxssantm1 kernel: Unable to handle kernel paging request > at virtual address f8d44080 > Dec 23 14:35:45 lxssantm1 kernel: printing eip: > Dec 23 14:35:45 lxssantm1 kernel: f8f47e43 > Dec 23 14:35:45 lxssantm1 kernel: *pde = 1fc86067 > Dec 23 14:35:45 lxssantm1 kernel: Oops: 0000 [#1] > Dec 23 14:35:45 lxssantm1 kernel: SMP > Dec 23 14:35:45 lxssantm1 kernel: Modules linked in: dm_round_robin > dm_multipath joydev st sr_mod ide_cd cdrom nvram edd speedstep_lib fre > q_table processor ipv6 sg intel_agp agpgart i2c_piix4 i2c_core evdev > tulip dm_mod reiserfs qla2300 qla2xxx scsi_transport_fc gdth aic7xxx > sd_mod scsi_mod > Dec 23 14:35:45 lxssantm1 kernel: CPU: 0 > Dec 23 14:35:45 lxssantm1 kernel: EIP: 0060:[<f8f47e43>] Not > tainted VLI > Dec 23 14:35:45 lxssantm1 kernel: EFLAGS: 00010247 (2.6.10-rc3) > Dec 23 14:35:45 lxssantm1 kernel: EIP is at trigger_event+0x3/0x10 > [dm_multipath] > Dec 23 14:35:45 lxssantm1 kernel: eax: f8d44080 ebx: f31787ec ecx: > dfe30000 edx: 00000216 > Dec 23 14:35:45 lxssantm1 kernel: esi: dfe30000 edi: f31787f0 ebp: > f3178760 esp: dfe33f60 > Dec 23 14:35:45 lxssantm1 multipathd: event checker startup : yellow > Dec 23 14:35:45 lxssantm1 kernel: ds: 007b es: 007b ss: 0068 > Dec 23 14:35:45 lxssantm1 kernel: Process events/0 (pid: 6, > threadinfo=dfe32000 task=dffefa60) > Dec 23 14:35:45 lxssantm1 kernel: Stack: c0128a38 dfe30020 f8f47e40 > 00000216 dfe32000 dfe3000c ffffffff ffffffff > Dec 23 14:35:45 lxssantm1 kernel: 00000001 00000000 c0116330 > 00010000 00000000 c18109c0 c1810060 00000000 > Dec 23 14:35:45 lxssantm1 kernel: 00000000 dffefa60 c0116330 > 00100100 00200200 dffefbc4 00000000 00000000 > Dec 23 14:35:45 lxssantm1 kernel: Call Trace: > Dec 23 14:35:45 lxssantm1 kernel: [<c0128a38>] worker_thread+0x1a8/0x230 > Dec 23 14:35:45 lxssantm1 kernel: [<f8f47e40>] trigger_event+0x0/0x10 > [dm_multipath] > Dec 23 14:35:45 lxssantm1 kernel: [<c0116330>] > default_wake_function+0x0/0x10 > Dec 23 14:35:45 lxssantm1 kernel: [<c0116330>] > default_wake_function+0x0/0x10 > Dec 23 14:35:45 lxssantm1 kernel: [<c0128890>] worker_thread+0x0/0x230 > Dec 23 14:35:45 lxssantm1 kernel: [<c012c666>] kthread+0x86/0xb0 > Dec 23 14:35:45 lxssantm1 kernel: [<c012c5e0>] kthread+0x0/0xb0 > Dec 23 14:35:45 lxssantm1 kernel: [<c01019b5>] > kernel_thread_helper+0x5/0x10 > Dec 23 14:35:45 lxssantm1 kernel: Code: 89 f0 e8 61 fc ff ff 8b 7e 38 85 > ff 75 96 8b 46 28 85 c0 74 96 8b 5e 2c 85 db 75 8f c7 44 24 04 01 > 00 00 00 eb 85 89 f6 8b 40 08 <8b> 00 e9 76 e8 d4 ff 8d b6 00 00 00 00 > 56 85 d2 53 89 ce 89 c3 > Dec 23 14:35:56 lxssantm1 multipathd: 8:64 : emc_clariion_checker: Path > healthy > > > Enable the Switch Port or plugging the cable has no effect. The > device /dev/yellow remains dead. > > No failover !!!! > > -- > > dm-devel@xxxxxxxxxx > https://www.redhat.com/mailman/listinfo/dm-devel -- christophe varoqui <christophe.varoqui@xxxxxxx>