2017-11-17 16:32 GMT+02:00 Lucas Stach l.stach@pengutronix.de:
Hi Kalle,
Am Freitag, den 17.11.2017, 16:27 +0200 schrieb Kalle Kankare:
Hi,
The regression was previously discussed on this mailing list last month (2017-10-04).
As in the previous discussion, this commit restricts the supported sound formats to S24_LE. This problem is experienced in our iMX6q based hardware with SGTL5000 sound chip. I.e.:
# arecord -l | grep 'card 0' -A2 card 0: imx6qba16sgtl50 [imx6q-ba16-sgtl5000], device 0: HiFi sgtl5000-0 [] Subdevices: 1/1 Subdevice #0: subdevice #0
In kernel 4.11.0:
# arecord --dump-hw-params -D hw:0,0 2>&1 | grep '^FORMAT' FORMAT: S16_LE
And in now in kernel 4.13.3:
# arecord --dump-hw-params -D hw:0,0 2>&1 | grep '^FORMAT' FORMAT: S24_LE
By reverting the mentioned commit 70d435ba1cd6, the format returned to S16_LE. However, I'm unsure if that is the proper way to solve this issue. I.e. does reverting the mentioned commit cause some other unwanted effect ?
What would be the best solution that would make the S16_LE format supported again for SGTL5000 ?
Please cherry-pick f9d4a398f121 (dmaengine: imx-sdma: Correct src_addr_widths and directions) from current Linus mainline tree.
Thank you. The commit fixes the issue.
Regards, Lucas