DMA mode and driver jax10

[Loïc Minier]

Hi,

on jax10 devices, a single physical Intel_SSD disk is split in two, ata1.00 and ata1.01, but the generic Ubuntu kernel only sees one.

This means that there's only enough space to host the OS on the first drive (/dev/sda, 1929692944 bytes), and the second drive for user data isn't visible.

Bye,

[Loïc Minier (lool)]

Amit, this patch isn't in the jaunty and intrepid trees; I checked the ubuntu-hardy debian/binary-custom.d/lpia/patchset/* patches and some parts of other patches are missing as well.

Please reapply the disk controller related patches in intrepid and jaunty ASAP; I'll bring the other topics up in another place.

[Loïc Minier (lool)]

Didn't test the patch but it builds

[Andy Whitcroft (apw)]

@Loic -- I have rebased the linux-lpia tree to the tip of the distro jaunty tree and applied the patch you suggest above for the jax10 problem. Could you test these and report back here. Kernels can be found in my personal PPA:

    https://launchpad.net/~apw/+archive/ppa

This package:

    linux-lpia - 2.6.28-2.3~jaunty1apw4

[Loïc Minier (lool)]

So I don't have a jax10 right now -- sigh, sorry -- I've asked others from the mobile team with a jax10 to test the lpia and i386 kernels in jaunty right now to see whether:
- they see 2 SSD drives 2 GB each
- the DMA mode is selected as UDMA 5 or similar on boot (please attach dmesg output!)

If the latter isn't the case, please try with the linux-lpia kernel in apw's PPA; thanks!

[Michael Casadevall (mcasadevall)]

After a lot of trouble, I got my jax10 working again, and booting the i386 2.6.28-8-generic kernel. Both SSD parts are detected. and the following UDMA messages are in the log:

[ 3.568479] ata1.00: configured for UDMA/66
[ 3.576475] ata1.01: configured for UDMA/66
[ 3.743437] scsi 0:0:0:0: Direct-Access ATA Intel_SSD 2008 PQ: 0 ANSI: 5

[Loïc Minier (lool)]

Michael, could you confirm the UDMA mode is the highest? (perhaps with hdparm) or perhaps try Andy's kernel and see whether you get a better mode

[Michael Casadevall (mcasadevall)]

Using apw's kernel, I got

[ 11.808387] ata1: PATA max UDMA/100 cmd 0x1f0 ctl 0x3f6 bmdma 0x1800 irq 14
[ 11.808509] ata2: PATA max UDMA/100 cmd 0x170 ctl 0x376 bmdma 0x1808 irq 15

Both devices are seen.

dmesg attached.

[Loïc Minier (lool)]

Andy, apparently the patch results in a different driver being selected for this device; what do you think is the best one?

Michael, lsmod?

[Michael Casadevall (mcasadevall)]

Sorry for the delay on this. lpia dmesg and lsmod attached. As far as I can tell, its just using the standard ATA driver with the normal lpia kernel; I'm not sure what I'm supposed to see beyond that.

[Michael Casadevall (mcasadevall)]

On the current lpia kernel:
Version 1.03c ------Sequential Output------ --Sequential Input- --Random-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
exodius 1G 4324 96 15681 19 7766 13 5015 95 33553 23 741.2 11
                    ------Sequential Create------ --------Random Create--------
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
                 16 11190 96 +++++ +++ 17303 96 7377 63 +++++ +++ 18149 100
exodius,1G,4324,96,15681,19,7766,13,5015,95,33553,23,741.2,11,16,11190,96,+++++,+++,17303,96,7377,63,+++++,+++,18149,100

On apw's kernel
Version 1.03c ------Sequential Output------ --Sequential Input- --Random-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
exodius 1G 4461 98 13590 16 11291 19 5225 99 33724 24 400.2 6
                    ------Sequential Create------ --------Random Create--------
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
                 16 11702 99 +++++ +++ 18586 99 7863 66 +++++ +++ 18321 100
exodius,1G,4461,98,13590,16,11291,19,5225,99,33724,24,400.2,6,16,11702,99,+++++,+++,18586,99,7863,66,+++++,+++,18321,100

There doesn't seem to be a massive difference between the two ...

[Michael Casadevall (mcasadevall)]

Talking with apw, the patch is unneeded since it does not increase performance. Marking invalid.