[#3902] spi_mmc bug

Submitted By: Phil Wilshire

Open Date

2008-02-17 01:01:42 Close Date

2009-03-02 05:45:53

Priority:

Medium Assignee:

Graf Yang

Status:

Closed Fixed In Release:

N/A

Found In Release:

snaps Release:

2008r1

Category:

N/A Board:

STAMP

Processor:

BF537 Silicon Revision:

0.2

Is this bug repeatable?:

Yes Resolution:

Out of Date

Uboot version or rev.:

Toolchain version or rev.:

2008r1

App binary format:

N/A

Summary: spi_mmc bug

Details:

spi_mmc_core.c sets the speed to SPI_CLOCK_MAX_HZ and not SPI_CLOCK_INIT_HZ

As a result the spi_mmc system does not seem to want to work.

Resetting it to SPI_CLOCK_INIT_HZ seems to make things work a little better.

Trivial Patch Attached

Phil Wilshire

Follow-ups

--- Vivi Li 2008-02-17 23:25:45

Now we are testing spi mmc with spi frame work and speed is set to

SPI_CLOCK_MAX_HZ. It works here.

Could you give a detailed process to reproduce your bug? Or your config file?

--- Phil Wilshire 2008-02-29 06:30:08

Hi Vivi,

The BF537 STAMP V1.3 with a Rubico SD card adaptor failed to detect

any of my 1G cards

(kodak, polariod, and brand X ?? ).

I do not have the actual board running at the moment so I cannot give you more

details.

Resetting the speed to SPI_CLOCK_INIT_HZ make the system detect the cards

initially.

(This is Probably a different bug)

In order to detect the card change I had to remove and re install the module

(It looks like PF5 was being held low and could not act as a card detect).

TO test this we also need to consider a test set up with concurrent aceess to

another device on the spi bus.

This would high light the major SPI-MMC problem we still have today.

NOte that work is going on to resolve this particular problem.

Thanks

Phil

--- Graf Yang 2008-03-06 00:02:16

We tried 512M MMC (Transend, Kingmax, Kingstone), 256M SD (two SanDisk cards),

1G SD (Transend), 2G SD (Transcend). All these cards can be detected, and we

needn't remove/reinstall spi_mmc module. We use latest trunk code and a Rubico

SD adaptor.

When make menuconfig, I selected MMC/SD card support --> MMC/SD for SPI

support and choiced the "SPI driver (SPI framework master)". I tried

seting the SPI clock to 400K and 20M HZ, both are OK.

As I have not such environment that with several devices concurrent access the

SPI bus, I can't duplicate the issue -- PF5 was being held low and could not act

as a card detect, at this moment.

--- Hans Eklund 2008-03-06 03:01:54

Yes. As phil pointed out the concurrency problem goes deep into the spi

framework and the spi_mmc driver needs to be rewritten slightly. Whenever a

"fix" is found I will try to incorporate it into the mainline mmc_spi

driver(another driver). I pointed out that issue earlier and can be found as a

tracker:

blackfin.uclinux.org/gf/project/uclinux-dist/tracker/?action=TrackerItemEdit&tracker_id=141&tracker_item_id=3713

I will post information there.

Files

Changes

Commits

Dependencies

Duplicates

Associations

Partner Forums

Partner Libraries

[#3902] spi_mmc bug