Alexander Stein wrote on 2011-05-04:
> Hello,
>
> I just encountered a problem on my ARMv5 (AT91SAM9263 based) board
> while using
> ads7843 or ads7846 (I replaced it). I tried reading
> /sys/class/hwmon/hwmon0/device/in0_input and got 0 most of the time.
> Sometimes the correct voltage was shown.
> After digging a bit in the code I found out that sample (and command
> for Tx) from the following struct gets DMA mapped by the underlying
> atmel_spi driver for SPI Rx-Buffer.
>
> struct ser_req {
> u8 ref_on;
> u8 command;
> u8 ref_off;
> u16 scratch;
> __be16 sample;
> struct spi_message msg;
> struct spi_transfer xfer[6];
> };
>
> The atmel_spi driver does proper dma_{,un}map_signgle on the Rx and Tx
> buffers. Now when the cache line for scratch is cleared and
> invalidated and the spi subsystem access msg (and maybe xfer) the
> cache line gets filled again, rendering the cache clear and
> invalidation before as useless.
> As the ARMv5 core has 32Byte cache lines I added "char dummy[25];"
> after sample, so the spi structs are in a different cache line.
> But this seem only to be a workaround as the cache lines might differ
> across each CPU.
> Somebody an idea how to fix this in general?
Take a look at the AD7877 Touchscreen driver.
To work around the same issue we use ____cacheline_aligned
/*
* DMA (thus cache coherency maintenance) requires the
* transfer buffers to live in their own cache lines.
*/
u16 conversion_data[AD7877_NR_SENSE] ____cacheline_aligned;
Alternatively you can kmalloc the transfer buffers.
Greetings,
Michael
--
Analog Devices GmbH Wilhelm-Wagenfeld-Str. 6 80807 Muenchen
Sitz der Gesellschaft: Muenchen; Registergericht: Muenchen HRB 40368; Geschaeftsfuehrer:Dr.Carsten Suckrow, Thomas Wessel, William A. Martin, Margaret Seif
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
