JMB363 false hotplug detections (was: ahci: AHCI and RAID mode SATA patch for Intel Cougar Point DeviceIDs)

From: Robert Hancock
Date: Sat Jan 16 2010 - 13:10:21 EST

On Sat, Jan 16, 2010 at 12:02 PM, Krzysztof Halasa <khc@xxxxxxxxx> wrote:
> Robert Hancock <hancockrwd@xxxxxxxxx> writes:
>> Hmm.. From those test results I really suspect some kind of hardware
>> fault. Could be a defective motherboard - I don't know if that chip
>> needs any terminating resistors on the motherboard for the SATA signal
>> lines or something, if so, could be they weren't installed properly..
> Unfortunately I can't find a JMB363 datasheet on the net, but there is
> a certain mb (p965t-a) schematic available.
> It seems JMB363 doesn't need terminators on SATA RX/TX lines, there is
> capacitative coupling only (10 nF in each line).
> The port in question (SATA#2) on my mb (P45 Neo2) uses pins 56 (RXP)
> 57 (RXN) and 60 (TXN) 61 (TXP). No visible irregularity, the traces look
> like they should, go straigt to the capacitors, and then to 0R R-packs
> and to the connector. It looks exactly the same for both ports. There is
> no short circuit past the capacitors (from the connector side). I'd say
> quite low probability that there is something wrong with these signals.
> It seems the chip uses extra 12k resistors for SATA (p965t-a calls the
> pins SJ_REXT[12]), pin 44 for port#1 and 55 for port#2. Both look sane.
> I will check the suspected connections with the machine powered off
> later.
> The RX and TX trace pairs go next to each other for up to 10 mm, could
> that be a problem at these frequencies? If so it would show up on
> all/many such boards certainly? Can't find any report.
> OTOH other people have similar problems with other boards: e.g.
> (unknown boards)
> The last one claims:
> this started after an upgrade to ubuntu 9.04 and is stll here after re-installing
> ubuntu 8.10.
> this was fixed by re-installing ubuntu 8.10 only using the kernal,
> 2.6.27-7-generic.
> I don't know if JMB36x is involved in this case, and how reliable the
> info is.
> Investigating as time permits.

Well, it is possible there is some kind of flaw in the JMB363 chip
itself that causes this problem. (Could be it happens in Windows too,
I don't think Windows drivers normally report these kinds of events
anywhere and if it never reached the point of actually deciding a
device was connected, you likely wouldn't notice.) I suppose we could
add a workaround in the driver to ignore hotplug events, but then real
hotplug events wouldn't get handled properly..

What revision does your JMB363 report in lspci? Mine shows:

03:00.0 SATA controller: JMicron Technologies, Inc. 20360/20363 Serial
ATA Controller (rev 03) (prog-if 01 [AHCI 1.0])

Tejun, any other ideas?
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at