So I've done quite a bit more research and have some more info if this helps anyone with ideas.

It seems that the mx31ads_pcmcia driver's .probe function is never called by the platform. The driver is created, but there are no devices to pair it with. Since there are no devices, it doesn't run probe. How are these devices supposed to be created? Where does that responsibility lie?

Thanks for any help you can provide....

Posted By davek on 12 Jul 2007 9:57 AM
It seems that the mx31ads_pcmcia driver''s .probe function is never called by the platform. The driver is created, but there are no devices to pair it with. Since there are no devices, it doesn''t run probe. How are these devices supposed to be created? Where does that responsibility lie?

In case you're using my patchset, I guess I'm the one to blame in this case.
The answer is simple: There are two strings:

drivers/pcmcia/mx31ads-pcmcia.h: 22
arch/arm/mach-mx3/devices.c: 587

those two have to be identical, otherwise the platform device is not probed.
I accidentally changed one of them, I guess.

Daniel

Hi zonque, thanks very much for your help. That solved that problem. I had temporarily added a platform_device_register( ) of my own, but this solution is more correct.

So it now detects the insertion of a CompactFlash memory card, but dump_cis says "no CIS present" for socket 0. Any suggestions?

Thanks so much!

If I startup the board with Lolo, and mount the flash card in the bootloader using:
mount fatfs /cf
and then proceed with the boot process, the card is properly detected and the proper windows are printed out with dump_cis.

The ide_cs driver properly loads, but the hotplug/udev scripts seem to be broken. More importantly I continually get the following message on the console:
[ 52.730000] hda: lost interrupt

And ide access to the device just seems to hang the kernel. Any suggestions?

A little bit more searching in the PCMCIA driver found the following line in mx31ads_pcmcia.c:

/*
* This really needs a better solution. The IRQ
* may or may not be claimed by the driver.
*/

Might this be a hint as to the solution?

I did the change in as above and am able to see the message:
[ 53.330000] pccard: PCMCIA card inserted into slot 0

But after this I dont see a hda message or hda device being created in /dev. I dont get a message that PCMCIA driver is initialised.
Also if i do a dump_cis i get a command not found. Do I need to run any script or the driver itself is not initialized correctly?

Hi Arjun.KV,

Make sure you have the PCMCIA user tool installed, they are in the packages pcmcia-cs and pcmciautils. Those are selectable in the ltib package manager.

In order for a flash card to work, you'll need to make sure you're building support for ide_cs into the kernel. Make sure you have PCMCIA IDE selected as built-in or module support. This will compile the legacy stuff.

That being said, I haven't been able to get it working. There seems to be a problem with interrupt support in this old driver. It detects the card, etc, but the I just get a bunch of hda: interrupt lost messages.

Dave

Hi,

I got the PCMCIA driver loaded and doing something, but somehow the I/O operations seem to fail. I always get zero Manufacturer and Device IDs when doing pccardctl info.

imx31:/mnt# pccardctl info
PRODID_1=""
PRODID_2=""
PRODID_3=""
PRODID_4=""
MANFID=0000,0000
FUNCID=255

I tried different Compact Flash cards inserted into the CF/PCMCIA slot, without success.

dmesg shows that the CF card is detected:
pccard: PCMCIA card inserted into slot 0

But somehow the ide-cs driver does not register any device. I think it is because of the MANFID and FUNCID having odd values.

Does anyone have an idea whats wrong ?

Thanks,
jdr

Hi jdr....

It could be a timing problem. I'm pretty sure that that's what I'm seeing.
On line 794 of mx31ads_pcmcia.c I changed:

code = PCMCIA_POR_PSHT(mx31ads_pcmcia_por_psht(speed_ns, clk_ns))
| PCMCIA_POR_PSST(mx31ads_pcmcia_por_psst(speed_ns, clk_ns))
| PCMCIA_POR_PSL(mx31ads_pcmcia_por_pslt(speed_ns, clk_ns));

to be:
code = PCMCIA_POR_PSHT(0)
| PCMCIA_POR_PSST(63)
| PCMCIA_POR_PSL(127);

Not sure if it's perfect, but it's better.

Hope that helps, let me know if it works for you.

Keep in mind that old versions of baseboard hardware have the CF interface signals messed up. Check PCN341 for more info:

http://www.logicpd.com/do..._iMX_LITEKIT_PCN.pdf

I have a 1007052 Rev 1 baseboard, so it appears that this problem has been fixed on the PCB that I have, correct?

If you can load a file from a cf card using lolo 2.3.5 you can be sure that's it's fixed. A workaround to the CF interface issue on old hardware was built into 2.3.4, so 2.3.4 works on the bad hardware and 2.3.5 works on the fixed hw.

Hi,

I am getting the following message when i try to use cf.

mx31# /etc/rc.d/init.d/pcmcia start
[ 51.960000] pcmcia_core: exports duplicate symbol dead_socket (owned by kernel)
[ 52.970000] mx31ads_pcmcia: Unknown symbol clk_get
[ 52.980000] mx31ads_pcmcia: Unknown symbol clk_get_rate
[ 51.030000] firmware_class: exports duplicate symbol release_firmware (owned by kernel)
[ 51.060000] pcmcia_core: exports duplicate symbol dead_socket (owned by kernel)
[ 51.090000] pcmcia: exports duplicate symbol cs_error (owned by kernel)
mx31# dump_cis
Socket 0:
no CIS present
mx31# cardctl info
PRODID_1=""
PRODID_2=""
PRODID_3=""
PRODID_4=""
MANFID=0000,0000
FUNCID=255

I did the timing changes in mx31ads_pcmcia.c as suggested.
Any reason why this is happening?

Paul, is there a way in LoLo to list CIS info from a card, or does it only know how to handle memory cards?

There isn't a built in command, although on previous products (a404/pxa270) there was a procedure to do this. It used the w, x, and remap commands to set up the interface and dump out the memory-mapped attribute area, I believe. I haven't worked the procedure out for the i.mx31 and I don't think that lolo itself looks at the card's cis (timing values are hard-coded).

Paul, if you figure out the method for the MX31 that would really help me prove where the problem is. Right now it looks like it's a baseboard or MX31 timing problem. I've sent the query off to Freescale and they have an FAE looking at it, but any insight you could provide would be appreciated.

Thanks
Dave

I will let you know.

In the meantime, the timing from the WinCE driver registry file might be helpful to you:

"NumSlots"=dword:1
"OffboardCtrlReg"=dword:16300000 ; only used if "NumSlots"=dword:2
; Set the strobe timing value according to CF spec
"PccStrobeLength"=dword:0000000F ; Strobe length maximum value is 0x7F
"PccStrobeSetupTime"=dword:0000000F ; Strobe setup time maximum value is 0x3F
"PccStrobeHoldTime"=dword:0000003F ; Strobe hold time maximum value is 0x3F

"SlotBIsrDll"="pcmcia_isrb.dll"
"SlotBIsrHandler"="ISRHandler"
"SlotBAttr"=dword:00000303 ; meet the minimum required in the CF spec
"SlotBCommon"=dword:00000203 ; if max timing is desired, ws will have to
"SlotBIO"=dword:00000284 ; be increased
"SlotBIrq"=dword:36


"SlotAIsrDll"="pcmcia_isra.dll"
"SlotAIsrHandler"="ISRHandler"
"SlotAAttr"=dword:00000303 ; meet the minimum required in the CF spec
"SlotACommon"=dword:00000203 ; if max timing is desired, ws will have to
"SlotAIO"=dword:00000284 ; be increased
"SlotAIrq"=dword:36

Hi,

This might sound silly, but what config settings did you use to generate the hdx devices.
[ 52.730000] hda: lost interrupt

Also when i try to mount the cf card in LoLo I get the following error:
losh> mount fatfs /cf
timeout resetting cf card
error: mount: failed to open (ATA)

Any suggestions would be great.

Arjun.KV,

The Pilot and Production i.MX31 kits need a little time to initialize the CF card. If you are using a script you can use the sleep <mseconds> command to give the kit time to do that.

sleep 3000;
mount fatfs /cf;

In some cases, having a sleep command between 2 mount commands helps. I've always had good success with just the one delay.

Thanks