R: ahci timeout

>
>>From : ndenev_at_gmail.com
>>
>>On 27 Dec, 2010, at 21:20 , Barbara wrote:
>>
>>> 
>>> As my old PATA hard disk was failing, I had to replace it with a new SATA 
>>> drive where I moved my FreeBSDs installations, as PATA drives are not 
easy 
>to 
>>> find these days.
>>> So I had to move one of my data drive from a VIA8237A SATA controller to 
>the 
>>> last free SATA slot on a Marvell 88SX6121 to make room for the new hd.
>>> The hd I moved was working perfectly when connected to the VIA controller.
>>> Now, with the Marvell I'm getting messages like the following twos while 
>using 
>>> the disk:
>>>    ahcich0: Timeout on slot 10
>>>    ahcich0: is 00000000 cs 3ffff800 ss 3ffffc00 rs 3ffffc00 tfd 50010040 
>serr 
>>> 00000000
>>> 
>>>    ahcich0: Timeout on slot 5
>>>    ahcich0: is 00000000 cs 00000180 ss 000001e0 rs 000001e0 tfd 50040040 
>serr 
>>> 00000000
>>> 
>>> This doesn't happen regularly. For example downloading from a slow 
website 
>on 
>>> it, so few kb/s, is ok.
>>> But if I copy files from the disk attacked to the Marvell controller to 
>>> another another disk, or for example run md5 on some files, it's very 
>likely to 
>>> happen.
>>> The process accessing the disk can not be killed even with -9, ^C does 
>>> nothing, and umount doesn't exit.
>>> If I'm copying files on it from another disk it can't be unmounted too as 
>the 
>>> unkillable process has it in use.
>>> On shutdown many disk doesn't get unmounted, so there are a lot of fsck 
on 
>>> boot, and on CURRENT (last built yesterday), FreeBSD enter debugger as it 
>fail 
>>> flushing disk caches.
>>> 
>>> Relevant part from dmesg:
>>> 
>>> atapci0: <Marvell 88SX6121 UDMA133 controller> port 0xdc00-0xdc07,0xd880-
>>> 0xd883,0xd800-0xd807,0xd480-0xd483,0xd400-0xd40f mem 0xfbdffc00-
0xfbdfffff 
>irq 
>>> 28 at device 0.0 on pci6
>>> ahci0: <Marvell 88SX6121 AHCI SATA controller> on atapci0
>>> ahci0: AHCI v1.00 with 2 3Gbps ports, Port Multiplier supported
>>> ahcich0: <AHCI channel> at channel 0 on ahci0
>>> ahcich1: <AHCI channel> at channel 1 on ahci0
>>> ata2: <ATA channel 0> on atapci0
>>> atapci1: <VIA 8237A SATA150 controller> port 0xbc00-0xbc07,0xb880-0xb883,
>>> 0xb800-0xb807,0xb480-0xb483,0xb400-0xb40f,0xb000-0xb0ff irq 21 at device 
>15.0 
>>> on pci0
>>> ata3: <ATA channel 0> on atapci1
>>> ata4: <ATA channel 1> on atapci1
>>> atapci2: <VIA 8237A UDMA133 controller> port 0x1f0-0x1f7,0x3f6,0x170-
0x177,
>>> 0x376,0xfc00-0xfc0f at device 15.1 on pci0
>>> ata0: <ATA channel 0> on atapci2
>>> ata1: <ATA channel 1> on atapci2
>>> 
>>> ada0 at ahcich0 bus 0 scbus0 target 0 lun 0
>>> ada0: <ST31000528AS CC44> ATA-8 SATA 2.x device
>>> ada0: 300.000MB/s transfers (SATA 2.x, UDMA6, PIO 8192bytes)
>>> ada0: Command Queueing enabled
>>> ada0: 953869MB (1953525168 512 byte sectors: 16H 63S/T 16383C)
>>> ada1 at ata3 bus 0 scbus3 target 0 lun 0
>>> ada1: <WDC WD2500KS-00MJB0 02.01C03> ATA-7 SATA 2.x device
>>> ada1: 150.000MB/s transfers (SATA 1.x, UDMA5, PIO 8192bytes)
>>> ada1: 238475MB (488397168 512 byte sectors: 16H 63S/T 16383C)
>>> ada2 at ata4 bus 0 scbus4 target 0 lun 0
>>> ada2: <ST3500320AS SD1A> ATA-8 SATA 1.x device
>>> ada2: 150.000MB/s transfers (SATA 1.x, UDMA5, PIO 8192bytes)
>>> ada2: 476940MB (976773168 512 byte sectors: 16H 63S/T 16383C)
>>> ada3 at ata0 bus 0 scbus5 target 0 lun 0
>>> ada3: <MAXTOR STM3160212A 3.AAJ> ATA-7 device
>>> ada3: 100.000MB/s transfers (UDMA5, PIO 8192bytes)
>>> ada3: 152627MB (312581808 512 byte sectors: 16H 63S/T 16383C)
>>> 
>>> _______________________________________________
>>> freebsd-current_at_freebsd.org mailing list
>>> http://lists.freebsd.org/mailman/listinfo/freebsd-current
>>> To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
>>
>>Just to add a "me too".
>>
>>I'm running -STABLE but have the same problems with Marvell 88SX6121 giving 
>"ahci timeout" messages.
>>
>>Regards,
>>Nikolay
>>
>
>Nikolay, thanks for the feedback, even if unfortunately it's negative...
>I see that in both 8-STABLE (8.2-PRERELEASE now) and 9.0-CURRENT, both 
rebuilt 
>no more than a week ago.
>I've also tried rebuilding the kernel with:
>   options         CAMDEBUG
>   options         CAM_DEBUG_BUS=0
>   options         CAM_DEBUG_TARGET=0
>   options         CAM_DEBUG_LUN=0
>   options         CAM_DEBUG_FLAGS=CAM_DEBUG_TRACE
>(0 is the bus/target/lun of the marvell controller/attached hd)
>and run a test computing md5 for about 60GB of files.
>Obviously, as the debug options where active, it run successfully without 
any 
>problems :)
>Is there any other info I should provide or any other test that I can do?
>
>Thanks
>Barbara
>

Maybe the attached disk has some problems which aren't handled if it's 
attached to a 88SE6121 controller?
>From what I can see, connecting it to the other internal sata controller, 
which is a VIA 8237A, or to an external PCIe Sil3132 controller using the siis 
driver, those timeouts aren't happening.
Anyway I see something which I don't understand.
I tried reading some files (~1 gb) from the same slice (1st one, ~200gb) while 
looking at gstat.
Some files are being read at >100mb/s some others at about 4 mb/s. It seems 
that a "zone" of the disk is very slow.
The disk is a Seagate 7200.12 and neither smartmontools nor SeaTools (Seagate 
diagnostic) report problems with it.
Could this explain the timeouts on the Marvell controller?
And BTW, does anyone knows which could be the problem with the disk?

Thanks
Barbara