Bug 12703 - Intel X25-E 32GB+MCP55_nvidia_sata+2.6.28=it not work together
Summary: Intel X25-E 32GB+MCP55_nvidia_sata+2.6.28=it not work together
Status: RESOLVED CODE_FIX
Alias: None
Product: IO/Storage
Classification: Unclassified
Component: Serial ATA (show other bugs)
Hardware: All Linux
: P1 high
Assignee: Tejun Heo
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-02-14 08:58 UTC by Dusan Pavlik
Modified: 2010-02-17 04:55 UTC (History)
2 users (show)

See Also:
Kernel Version: 2.6.28.4 - mandriva 2009.1beta
Subsystem:
Regression: No
Bisected commit-id:


Attachments
Boot from live cd (29.01 KB, text/plain)
2009-02-15 00:10 UTC, Dusan Pavlik
Details
not booting with new kernel _1 (330.19 KB, image/jpeg)
2009-02-16 07:51 UTC, Dusan_pavlik_2
Details
not booting with new kernel _2 (346.68 KB, image/jpeg)
2009-02-16 08:03 UTC, Dusan_pavlik_2
Details
opensuse 11.1 -mesage (45.09 KB, application/octet-stream)
2009-03-01 00:34 UTC, Dusan Pavlik
Details
dmesg from 2.6.29.1 (35.98 KB, text/plain)
2009-04-09 14:14 UTC, Per Wigren
Details
nv-deb-long.patch (463 bytes, patch)
2009-04-10 19:23 UTC, Tejun Heo
Details | Diff
2.6.30-rc6 dmesg patched nohrst (40.69 KB, application/octet-stream)
2009-05-20 02:05 UTC, David Lang
Details
dmesg 2.6.30-rc6 patched (40.04 KB, application/octet-stream)
2009-05-20 02:05 UTC, David Lang
Details
2.6.30-rc6 kernel config (43.44 KB, application/octet-stream)
2009-05-20 02:06 UTC, David Lang
Details
dmesg with force 1.5 (40.11 KB, application/octet-stream)
2009-05-22 01:19 UTC, David Lang
Details
no-ipm-on-resume.patch (490 bytes, patch)
2009-05-22 01:28 UTC, Tejun Heo
Details | Diff
dmesg with second patch added (40.01 KB, application/octet-stream)
2009-05-22 02:00 UTC, David Lang
Details
nv-hardreset-only-on-probing.patch (4.36 KB, patch)
2009-05-31 01:55 UTC, Tejun Heo
Details | Diff
nv-hardreset-only-on-probing.patch (4.36 KB, patch)
2009-05-31 01:56 UTC, Tejun Heo
Details | Diff
dmesg with the patch, it appears to work (44.02 KB, application/octet-stream)
2009-06-02 22:03 UTC, David Lang
Details
updated dmesg with detatch and reconnect (45.88 KB, application/octet-stream)
2009-06-03 03:36 UTC, David Lang
Details

Description Dusan Pavlik 2009-02-14 08:58:04 UTC
Latest working kernel version:MDV 2.6.24.7, Fed10 2.6.27.5
Earliest failing kernel version:MDV 2.6.27.0, MDV 2.6.28.4, Fed11(alpha)2.6.29
Distribution: Mandriva, Fedora
Hardware Environment:Intel X25-E 32GB, MCP55_nvidia_sata
Software Environment: Standard distribution core software (kernel, bash, ...)
Problem Description: If I us Intel X25-E 32GB with the latest stable kernel is not working MDV 2.6.28.4. This ssd disk is not visible for system (logs, mesages,/dev/sdx, ...). If I using standard sata disc (samsung, western)-its work

if it work in mandriva 2008.1
lsmod
forcedeth              54156  0
ide_cd                 43552  0
button                 11808  0
thermal                19616  0
processor              37832  2 powernow_k8,thermal
i2c_nforce2            10240  0
i2c_core               25600  2 nvidia,i2c_nforce2
evdev                  14976  3
sg                     39192  0
ide_disk               18816  0
amd74xx                14104  0 [permanent]
ide_core              123928  3 ide_cd,ide_disk,amd74xx
sata_nv                29704  2
libata                155952  1 sata_nv
sd_mod                 31872  3
scsi_mod              157880  3 sg,libata,sd_mod
ext3                  137488  1
jbd                    51752  1 ext3


lspci
00:04.0 IDE interface: nVidia Corporation MCP55 IDE (rev a1)
00:05.0 IDE interface: nVidia Corporation MCP55 SATA Controller (rev a3)
00:05.1 IDE interface: nVidia Corporation MCP55 SATA Controller (rev a3)
00:05.2 IDE interface: nVidia Corporation MCP55 SATA Controller (rev a3)

lshw
 *-ide:1
          description: IDE interface
          product: MCP55 SATA Controller
          vendor: nVidia Corporation
          physical id: 5
          bus info: pci@0000:00:05.0
          logical name: scsi0
          version: a3
          width: 32 bits
          clock: 66MHz
          capabilities: ide bus_master cap_list emulated
          configuration: driver=sata_nv latency=0 maxlatency=1 mingnt=3 module=sata_nv
        *-disk UNCLAIMED
             description: ATA Disk
             product: SSDSA2SH032G1GN
             physical id: 0.0.0
             bus info: scsi@0:0.0.0
             version: 045C
             serial: CVEM849300LG032HGN
             configuration: ansiversion=5

Steps to reproduce:
If I install in Mandriva 2008.1 kernel 2.6.28.4 from Mandriva 2009.1 beta the the mesiges from the post is:

open /proc/scsi/scsi: no suchfile or direktory
Loading amd74xx module
amd74xx: Unknow symbol ide_pci_init_one
amd74xx: Unknow symbol ide_timing_compute
amd74xx: Unknow symbol ide_timing_merge
amd74xx: Unknow symbol ide_pci_remove
amd74xx: Unknow symbol ide_pci_suspend
amd74xx: Unknow symbol ide_pci_resume
amd74xx: Unknow symbol ide_pci_clk
Fatal: Error inserting amd74xx .....
Trying....
...  
...
...
Booting has failed
Comment 1 Dusan Pavlik 2009-02-15 00:10:08 UTC
Created attachment 20247 [details]
Boot from live cd

When it not work. Mandriva 2009

The /dev/sdc is not created. (is pluget on sata6). But is isn´t work when is pluget in SATA1.  Is not important.
Comment 2 Tejun Heo 2009-02-15 18:31:39 UTC
Hello, Dusan.  I'm having a difficult time understanding your last sentence.  So, the ssd works if it's connected to ata1 but doesn't when it's plugged into ata6?
Comment 3 Dusan_pavlik_2 2009-02-16 07:51:00 UTC
Created attachment 20259 [details]
not booting with new kernel _1

Is not work with every sata ports (SATA1, SATA2,....)

I sending new experiment with it. I´m installing MAndriva2008.1(when it´s work)
and I make upgrade to 2009.1beta(kernel mdv2.6.28.4). Everything is ok when I booting on old kernel (mdv2.6.24.7). When I booting in new kernel mdv2.6.28.4 it don´t booting  and is only writing which is in attachment.

Ps. sorry for my new account and English
Comment 4 Dusan_pavlik_2 2009-02-16 08:03:28 UTC
Created attachment 20260 [details]
not booting with new kernel _2
Comment 5 Tejun Heo 2009-02-18 22:40:55 UTC
Nah.. My English isn't that good either.  Don't worry about it.  Hmm... you're using IDE drivers and I'm not too familiar with mandravia.  Can you please give a shot at openSUSE 11.1 live CD?  Pressing alt-f4 will give you the kernel message console once the kernel is loaded.
Comment 6 Dusan_pavlik_2 2009-02-21 23:04:41 UTC
It may be the same problem. 

http://bugs.archlinux.org/task/12361
http://bugzilla.kernel.org/show_bug.cgi?id=12176

But why other standard sata disk is working? With Mandriva kernel 2.6.28.4.

I send new experiences when I will be home 28.2.2009.
Comment 7 Tejun Heo 2009-02-24 22:24:18 UTC
Those two bug reports are different ones and happens no matter which device is attached.  Yours seems to be PHY compatibility problem between the SSD and the controller.  Does specifying "libata.force=1.5Gbps" help?
Comment 8 Dusan Pavlik 2009-03-01 00:34:50 UTC
Created attachment 20391 [details]
opensuse 11.1 -mesage

Attachment 4 [details].
I use Opensuse live cd 11.1. It working like as in fedora 10.

"libata.force=1.5Gbps" is not working
Comment 9 Tejun Heo 2009-03-01 04:45:30 UTC
Dusan, can you please elaborate a bit?  From the posted log from openSUSE 11.1, it looks like everything is fine, right?  Also, what do you mean by "libata.force=1.5Gbps" is not working?  The linked attachment doesn't seem to be related?
Comment 10 Dusan Pavlik 2009-03-01 11:09:13 UTC
libata.force=1.5Gbps is not working with kernel MDV 2.6.28.4, MDV 2.6.29 rc6, Fed11(alpha)2.6.29. (I´m try which you writing in Comment 7)

In Opensuse live is kernel 2.6.27.7=its work
In Fedora 10 2.6.27.5 =its work
In Mandriva kernel 2.6.24 =its work

Its may problem with patch for kernel 2.6.28 (its may be old because MDV with 2.6.27.0 is doesn't work)

http://git.kernel.org/?p=linux%2Fkernel%2Fgit%2Fstable%2Flinux-2.6.28.y.git&a=search&h=HEAD&st=commit&s=amd74xx

http://git.kernel.org/?p=linux%2Fkernel%2Fgit%2Fstable%2Flinux-2.6.28.y.git&a=search&h=HEAD&st=commit&s=MCP55

Thanks.
Comment 11 Tejun Heo 2009-03-01 23:29:57 UTC
Dusan, I'm sorry but I'm still having problem understanding what you're reporting.  According to the log posted in comment#8, the intel SSD is working fine on ata8 at 3Gbps without libata.force=1.5Gbps, so I take it that the SSD works fine on both openSUSE 11.1 without any parameter, right?

I don't really understand what you mean by "libata.force=1.5Gbps is not working".  Do you mean that without it the detection is fine but with it detection doesn't work?  Or the parameter libata.force=1.5Gbps isn't effective at all?

So, to sum it up,

* openSUSE 2.6.27.7, fedora 2.6.27.5 and Mandravia 2.6.25 work fine.
* MDV 2.6.28.4 and 2.6.29rc6 and Fedora 2.6.28 don't work.

Am I getting it right?
Comment 12 Dusan_pavlik_2 2009-03-03 11:25:02 UTC
Yes, you are right.

In first report I writing
"Latest working kernel version:MDV 2.6.24.7, Fed10 2.6.27.5
Earliest failing kernel version:MDV 2.6.27.0, MDV 2.6.28.4, Fed11(alpha)2.6.29"

I´m add in grub parameter "libata.force=1.5Gbps" and is not working for kernels MDV 2.6.27.0, MDV 2.6.28.4, Fed11(alpha)2.6.29. (MDV is short cut for mandriva)

Do you mean I must buy new motherboard with chipset from AMD or Intel processor? I´ts really not good. I have AMD processor with TDP 35W(I have passive cooling for this processor).
Comment 13 Tejun Heo 2009-03-03 17:26:38 UTC
Ah.. okay, so libata.force doesn't make any difference and later kernels don't work.  And no, you shouldn't need to buy a new board.  We're having some trouble with nv device detection.

Are you comfortable with building your own kernel?  If so, can you please give a shot at 2.6.28.7?  There have been several nv reset related updates.

Thanks.
Comment 14 Dusan Pavlik 2009-03-08 04:47:47 UTC
I compiling kernel 2.6.28.7 (from kernel.org - make xconfig, make, make modules_install, make install) and instaling kernel-linus-2.6.29.0 rc7.1(from mandriva -rpm) its looking good when is starting initrd.img. (its very fast but I see it make the sda device)

But when loading the kernel modules the system is not start. It may be problem with "nash" script because when I install this kernels its working  with 100% load. ( make install is stoping with errors)
I wait for updait nash  and I report if work Intel SSD with this kernels.

Thanks. 
Comment 15 Tejun Heo 2009-03-08 16:54:26 UTC
If you include all the drivers and filesystem for root fs in the kernel proper (y instead of m), you won't need initrd or nash.  It's usually easier to test that way.
Comment 16 Per Wigren 2009-04-09 14:01:56 UTC
FYI, this problem still exists in 2.6.29.1. I still can't use my Intel X-25-E in kernels newer than 2.6.28.3 :(
Comment 17 Tejun Heo 2009-04-09 14:04:50 UTC
Can you please post failing boot log from 2.6.29.1?  We presumably should have all nv detection problems fixed now.
Comment 18 Per Wigren 2009-04-09 14:14:23 UTC
Created attachment 20906 [details]
dmesg from 2.6.29.1
Comment 19 Per Wigren 2009-04-09 14:18:00 UTC
As I have cloned the SSD to a standard SATA disk temporarily until I can get the SSD, the boot didn't fail, but there is no trace of the Intel SSD disk.

The Intel X-25-E is on SATA-port 1 and I have a 300 GB Maxtor on SATA port 2 which I booted from, which is seen as /dev/sda. (an iSCSI disk on /dev/sdb get added later on in the init scripts)
Comment 20 Per Wigren 2009-04-09 14:32:15 UTC
This is the related dmesg lines from 2.6.28.3:

ata6: SATA max UDMA/133 cmd 0xc000 ctl 0xbc00 bmdma 0xb808 irq 23
ata6: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
ata6.00: ATA-7: SSDSA2SH032G1GN INTEL, 045C8621, max UDMA/133
ata6.00: 62500000 sectors, multi 1: LBA48 NCQ (depth 31)
ata6.00: configured for UDMA/133
sd 6:0:0:0: [sdf] 62500000 512-byte hardware sectors: (32.0 GB/29.8 GiB)
sd 6:0:0:0: [sdf] Write Protect is off
sd 6:0:0:0: [sdf] Mode Sense: 00 3a 00 00
sd 6:0:0:0: [sdf] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
sd 6:0:0:0: [sdf] Attached SCSI disk
sd 6:0:0:0: Attached scsi generic sg5 type 0

(I had more disks connected then)
Comment 21 Tejun Heo 2009-04-09 18:04:55 UTC
Does "libata.force=6:nohrst" make nay difference?
Comment 22 Per Wigren 2009-04-09 23:25:49 UTC
Yes! With that option it works!  Is it safe to use the disk with that option until a proper fix is available?
Comment 23 Tejun Heo 2009-04-10 02:02:09 UTC
Yeap, it's safe to run with the option.  Hmm... I wonder why it isn't working without the option.  :-(  I'll try to think of something.
Comment 24 Tejun Heo 2009-04-10 19:23:48 UTC
Created attachment 20932 [details]
nv-deb-long.patch

Can you please give a shot at the attached patch and see whether there's any difference?  Thanks.
Comment 25 Per Wigren 2009-04-11 08:55:13 UTC
I will give it a try after Easter. I don't have time at the moment. Thanks!
Comment 26 Dusan Pavlik 2009-04-12 10:41:24 UTC
This parameter "libata.force=6:nohrst" for installation CD mandriva 2009.1 RC2 (kernel 2.6.29) is not work. System is doesn't see the Intel disk. 

Do you mean problem is only with chip-set driver? Or problem is between new kernel sata system and Intel SSD disk? Because I can´t try this disk in other PC with other chip-set.

Nash script problem is still her for upgrading new kernel. (It may be problem with upgrade from mdv2008.1) https://qa.mandriva.com/show_bug.cgi?id=48790


Thanks.
Comment 27 Tejun Heo 2009-04-14 03:50:25 UTC
Dusan, you'll probably need to do something inside initrd image.  It differs between distros and I don't know much about mandriva.

Yes, it's highly likely to be chipset-specific.
Comment 28 David Lang 2009-05-18 16:11:21 UTC
I am seeing the same problem on 2.6.29.3 I have not yet tried the parameter or patch, I will try to do so later today.
Comment 29 David Lang 2009-05-18 16:52:08 UTC
adding libata.force=1:nohrst lets it detect the intel drive on the first sata connector.

In my case I am using a monolithic kernel.
Comment 30 David Lang 2009-05-18 17:38:39 UTC
applying the patch  (nv-deb-long.patch)   does not help
Comment 31 Tejun Heo 2009-05-19 06:57:30 UTC
David, can you please post kernel boot log with the patch applied?

Thanks.
Comment 32 Tejun Heo 2009-05-19 06:58:04 UTC
And boot log with nohrst workaround.
Comment 33 David Lang 2009-05-20 02:05:08 UTC
Created attachment 21442 [details]
2.6.30-rc6 dmesg patched nohrst
Comment 34 David Lang 2009-05-20 02:05:51 UTC
Created attachment 21443 [details]
dmesg 2.6.30-rc6 patched
Comment 35 David Lang 2009-05-20 02:06:29 UTC
Created attachment 21444 [details]
2.6.30-rc6 kernel config
Comment 36 Tejun Heo 2009-05-21 02:12:48 UTC
Strange.  Interface is active but there's no sign of device (DET is zero).  I don't have much idea what's going on.  Does "libata.force=2:1.5Gbps" make any difference?

Thanks.
Comment 37 David Lang 2009-05-21 18:30:03 UTC
adding libata.force=2:1.5Gbps doesn't make any difference to the link down message, it still shows SStatus 100

when you say 'DET is zero', what are you looking at to see that?
Comment 38 Tejun Heo 2009-05-22 01:14:53 UTC
The SStatus has three four-bit fields IPM, SPD and DET, so SStatus 0x123 indicates IPM=1, SPD=2, DET=3 which means interface is active at Gen2 speed with device present and Phy communication established.  The port the ssd is attached to is reporting 0x100 which says link is powered up but nothing seems to be attached to the port.  Unfortunately, I don't have much idea how to proceed on this.  Ergh...

Can you please post the boot log with 1.5Gbps parameter just in case?

Thanks.
Comment 39 David Lang 2009-05-22 01:19:15 UTC
Created attachment 21476 [details]
dmesg with force 1.5
Comment 40 Tejun Heo 2009-05-22 01:28:02 UTC
Created attachment 21477 [details]
no-ipm-on-resume.patch

Can you please give a shot at this patch?
Comment 41 David Lang 2009-05-22 01:52:36 UTC
no difference
Comment 42 Tejun Heo 2009-05-22 01:55:37 UTC
Can you please attach boot log?  Even if it doesn't look useful now, it might turn out to be later, so...
Comment 43 David Lang 2009-05-22 02:00:35 UTC
Created attachment 21478 [details]
dmesg with second patch added

one interesting thing I noticed is that the order of the ports reporting has changed since the initial runs, it was in numerical order, now it's not. is this the result of the prior patch?
Comment 44 Tejun Heo 2009-05-22 05:26:57 UTC
No, it's not.  The order is non-deterministic depending on the timing of things.  If you unplug and replug the drive after detection failed, does it get detected?	Release Date: 06/26/2007
Comment 45 David Lang 2009-05-22 19:05:55 UTC
it is detected on a replug

from dmesg

ata2: exception Emask 0x10 SAct 0x0 SErr 0x150000 action 0xe frozen
ata2: SError: { PHYRdyChg CommWake Dispar }
ata2: hard resetting link
ata2: SATA link up 3.0 Gbps (SStatus 123 SControl 0)
ata2.00: ATA-7: SSDSA2SH032G1GN INTEL, 045C8621, max UDMA/133
ata2.00: 62500000 sectors, multi 1: LBA48 NCQ (depth 31)
ata2.00: configured for UDMA/133
ata2: EH complete
scsi 1:0:0:0: Direct-Access     ATA      SSDSA2SH032G1GN  045C PQ: 0 ANSI: 5
sd 1:0:0:0: Attached scsi generic sg2 type 0
sd 1:0:0:0: [sdb] 62500000 512-byte hardware sectors: (32.0 GB/29.8 GiB)
sd 1:0:0:0: [sdb] Write Protect is off
sd 1:0:0:0: [sdb] Mode Sense: 00 3a 00 00
sd 1:0:0:0: [sdb] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
 sdb: unknown partition table
sd 1:0:0:0: [sdb] Attached SCSI disk
Comment 46 David Lang 2009-05-28 21:24:16 UTC
anything else for me to try?
Comment 47 Tejun Heo 2009-05-31 01:55:15 UTC
Created attachment 21646 [details]
nv-hardreset-only-on-probing.patch

Can you please try this patch?  Thanks.
Comment 48 Tejun Heo 2009-05-31 01:56:41 UTC
Created attachment 21648 [details]
nv-hardreset-only-on-probing.patch

Oops, inverted condition.  Please test this one.  Thanks.
Comment 49 David Lang 2009-06-02 22:03:18 UTC
Created attachment 21716 [details]
dmesg with the patch, it appears to work
Comment 50 Tejun Heo 2009-06-03 00:58:09 UTC
Can you please try the followings?

1. Boot with the patch applied.
2. Disconnect the SSD.
3. echo - - - > /sys/class/scsi_host/host2/scan
4. Wait for the device to be detached.
5. Replug the SSD.
6. echo - - - > /sys/class/scsi_host/host2/scan
7. Wait for the device to be attached.
8. Attach dmesg.

Thanks.
Comment 51 David Lang 2009-06-03 03:36:45 UTC
Created attachment 21722 [details]
updated dmesg with detatch and reconnect
Comment 52 Dusan Pavlik 2009-06-26 13:22:26 UTC
This patch from comment #48 is good? It will be on kernel 2.6.31? (this kernel will be in mandriva 2010.0).

Thanks.
Comment 53 Dusan Pavlik 2009-06-26 14:55:16 UTC
OK I see it on change log on http://kernel.org/pub/linux/kernel/v2.6/testing/ChangeLog-2.6.31-rc1.
If I want it using for my SSD, I must add any parameter to grub or anything else?
Comment 54 David Lang 2009-06-26 20:01:04 UTC
no, with the patch no grub changes are needed
Comment 55 Tejun Heo 2010-02-17 04:55:19 UTC
Resolving as FIXED.

Note You need to log in before you can comment on or make changes to this bug.