2004-06-07 05:43:01

by Mikael Starvik

[permalink] [raw]
Subject: RE: [PATCH] CRIS architecture update

>It will probably fail on devices needing write access because
>e100_dma_begin() always passes e100_read_command as
>argument to e100_start_dma().

e100_read_command is a variable that is either 1 for reads
or 0 for writes. This variable is set in e100_dma_write and
e100_dma_read. As long as the framework calls dma_write or
dma_read before it calls dma_begin it will work just fine.

The driver has been tested with a ext2 filesystem on a disk
(and the files are still there after a sync and reboot).

/Mikael