Tejun did come up with a patch that makes this work, I haven't heard from
him in a couple of days so I don't know if he considers it good enough to
go in or not.
it would be very nice to have this expensive SSD usable without having to
patch the kernel or run an old kernel ;-)
I verified that 2.6.18 from centos 5.3 is able to see the SSD on this
hardware, but 2.6.28-13 (debian) and 2.6.29-[13] and 2.6.30-rc[67] cannot.
for those who have seen my other problem report on missing cores, this
SATA problem happens on systems that have no problem at all with the CPUs.
the bugzilla for this is http://bugzilla.kernel.org/show_bug.cgi?id=12703
I will have time for testing this tomorrow, but then will be out of town
for a couple of days.
David Lang
On Tue, 19 May 2009, Tejun Heo wrote:
> [email protected] wrote:
>> according to but 12703 ( http://bugzilla.kernel.org/show_bug.cgi?id=12703 )
>> this worked with older kernels, but with 2.6.29.[13] and 2.6.30-rc6
>>
>> according to this ticket it did work with older vendor kernels, but it's
>> not clear which ones (and no older vanilla kernels)
>>
>> I have verified that this drive does work on a PCI controller card in
>> the same machine (with the same cables), so this is definantly a
>> controller/kernel issue
>>
>> booting with libata.force=#:nohrst (with # being the SATA controller
>> that the SSD is on) does allow it to be detected, but the patch listed
>> in the ticket
>>
>> diff --git a/drivers/ata/sata_nv.c b/drivers/ata/sata_nv.c
>> index 6cda12b..6673b1d 100644
>> --- a/drivers/ata/sata_nv.c
>> +++ b/drivers/ata/sata_nv.c
>> @@ -1565,7 +1565,7 @@ static int nv_noclassify_hardreset(struct ata_link
>> *link, unsigned int *class,
>> bool online;
>> int rc;
>>
>> - rc = sata_link_hardreset(link, sata_deb_timing_hotplug, deadline,
>> + rc = sata_link_hardreset(link, sata_deb_timing_long, deadline,
>> &online, NULL);
>> return online ? -EAGAIN : rc;
>> }
>>
>> does not work.
>>
>> Tejun Heo was working on this back in April, but it looks like the
>> troubleshooting petered out.
>>
>> one issue that I did notice, in the dmesg I see
>>
>> $ grep SATA dmesg
>> ata1: SATA max UDMA/133 cmd 0xd480 ctl 0xd400 bmdma 0xcc00 irq 11
>> ata2: SATA max UDMA/133 cmd 0xd080 ctl 0xd000 bmdma 0xcc08 irq 11
>> ata1: SATA link down (SStatus 100 SControl 300)
>> ata2: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
>> ata3: SATA max UDMA/133 cmd 0xc880 ctl 0xc800 bmdma 0xc080 irq 5
>> ata4: SATA max UDMA/133 cmd 0xc480 ctl 0xc400 bmdma 0xc088 irq 5
>> ata3: SATA link down (SStatus 0 SControl 300)
>> ata4: SATA link down (SStatus 0 SControl 300)
>> ata5: SATA max UDMA/133 cmd 0xc000 ctl 0xbc00 bmdma 0xb480 irq 10
>> ata6: SATA max UDMA/133 cmd 0xb880 ctl 0xb800 bmdma 0xb488 irq 10
>> ata5: SATA link down (SStatus 0 SControl 300)
>> ata6: SATA link down (SStatus 0 SControl 300)
>>
>> note the ata1 line. it detects something different than the links where
>> there really is nothing plugged in, but it doesn't come up
>>
>> booting with nohrst results in SStatus 123
>
> Great, a new tester. :-)
>
> Let's track it on the bz.
>
> Thanks.
>
>
Hello,
[email protected] wrote:
> Tejun did come up with a patch that makes this work, I haven't heard
> from him in a couple of days so I don't know if he considers it good
> enough to go in or not.
>
> it would be very nice to have this expensive SSD usable without having
> to patch the kernel or run an old kernel ;-)
>
> I verified that 2.6.18 from centos 5.3 is able to see the SSD on this
> hardware, but 2.6.28-13 (debian) and 2.6.29-[13] and 2.6.30-rc[67] cannot.
I'll clean it up a bit and push upstream for 2.6.31. I'm a little bit
afraid that the behavior change is not minor and has potential of
breaking other configurations. So, for the time being, 2.6.30 would
require libata.force=nohrst as workaround.
Thanks.
--
tejun