On Thu, 2 Apr 2009, Tejun Heo wrote:
Grant Grundler wrote:
I could drop down to 2.6.26 and rebuild my kernel with support for that
controller - I hadn't done this before because I cannot then use my DVB-S
card, but... I could live without that for the test.
If someone could comment on 2.6.28 or 2.6.29, you could avoid running
this test.
Hmmm... I don't really remember changing much, so it should be fine.
Can you please give a shot at 2.6.26 anyway? Also, kernel logs with
JMB controller would be helpful too as ahci tends to better show what's
really going on and failing during reset.
I tried the JMB controller again with 2.6.26, and with drives 1,2,3
connected. This was different to what had happened previously - there are
two lights on the bridgeboard, green and red. Most of the time they're
both on. When the JMB controller was connected (even in the BIOS screen)
these lights flashed on and off about once a second. Occassionally they'd
settle into the on state. The log of booting into 2.6.26 is here:
http://usenet.gerph.org/SATA/sata-123-2.6.26-jmb-kern.log
and mostly consists of:
Apr 3 20:33:43 buttercup kernel: ata1: exception Emask 0x10 SAct 0x0 SErr 0x4000000 action 0xe frozen
Apr 3 20:33:43 buttercup kernel: ata1: irq_stat 0x00000040, connection status changed
Apr 3 20:33:43 buttercup kernel: ata1: SError: { DevExch }
Apr 3 20:33:43 buttercup kernel: ata1: hard resetting link
Apr 3 20:33:44 buttercup kernel: ata1: SATA link down (SStatus 0 SControl 300)
Apr 3 20:33:44 buttercup kernel: ata1: EH complete
repeated over and over whilst the light was flashing (not in time with
it).
An interesting (?) thing here is that at 20:34:27 in the log the light
stayed on for a bit and we managed to read details from the drives. Then
the light went back to flashing regularly again.
I also booted into 2.6.29 with this configuration (but with drives 3,4,5)
and got a similar effect, except for the constant light:
http://usenet.gerph.org/SATA/sata-345-2.6.29-jmb-kern.log
I don't think it's particularly surprising that this exhibits the same
results, given that even during the BIOS the light flashes in the same
way.
I removed and re-fitted the JMB card as part of these tests, in case it
was not seated properly in the slot (unlikely, as it was detected, but
just to be sure, I did this anyhow). I also exchanged the eSATA cables to
try to eliminate them. No change in behaviour was observed.
Having tried this, I reinserted the SiI card and rebuilt 2.6.26 with
support for it (and increased the debug buffer). Whilst in the BIOS and
after booting normally, the bridge board lights are almost constantly on.
They only seem to go off when the link is reset. Booting with drives
1,2,3,4 connected (which should be a failure, or only partially detect the
drives according to the 2.6.29 results) gave me:
http://usenet.gerph.org/SATA/sata-1234-2.6.26-sil-kern.log
which did many resets during the initialisation and never detected any
drives. Turning the external box off and on again did not cause it to
re-detect the drives (as if it was just completely ignoring the interface
now).
Repeating the test with drives 3,4,5 and the 2.6.26 kernel and the sil
card gave me the following log:
http://usenet.gerph.org/SATA/sata-345-2.6.26-sil-kern.log
This managed to identify 2 of the 3 drives during initialisation. When the
box was turned off and on again it only detected 2 of the 3.
I've replaced the PSU once already - it was replaced on 6th Feb, with a
Sumvision 450W 20+4pin SATA PSU. Those 5 drives + the bridge board and fan
are the only things being powered by that PSU. I should have thought that
even during drive initialisation 5 drives wouldn't exceed 450W. Anyhow, this
was one of my first thoughts. I could get another PSU (I think I might be
able to find one around here), but having replaced it so recently, I'm
uncomfortable doing so.
Is a 450W PSU going to be able to supply enough power for those drives ?
It's not just about how many watts. Both 12v and 5v "rails" need to
provide enough power. But if this worked before, I would assume the
replacement is working too.
And in general, yes, a 450W should be plenty for 5 drives plus PMP
board. Especially if the OS is "staggering" the spinup so only one
or two drives are spinning up at the same time. I know I've had to
fix "staggered spinup" in the past (2.6.18) but don't recall if it's
fixed in current 2.6.28 or .29 kernels.
libata never had proper provision for staggered spin up. If it works,
it's just because they're being probed sequentially. :-)
Anyways, 3132 + 3726 should work. Both 3726 and 4726 pretty much
require its first port to be always occupied (or it's fake config
drive jumps arounds and freaks out) and hotplug is often flaky but
other than that it should generally work.
So my using ports 3,4,5 only probably isn't all that useful for it ?
Even still, I'm not sure that the behaviour that I'm seeing is down to a
driver issue now. I'd appreciate any other thoughts based on the extra
info I've given. I may just bite the bullet and try a replacement bridge
board and see if that helps.
--
Gerph <http://gerph.org/>
[ All information, speculation, opinion or data within, or attached to,
this email is private and confidential. Such content may not be
disclosed to third parties, or a public forum, without explicit
permission being granted. ]
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
