diff options
| author | Jeff Garzik | 2007-02-26 07:26:06 +0100 |
|---|---|---|
| committer | Jeff Garzik | 2007-04-28 20:15:55 +0200 |
| commit | d88184fb2348a50f7c34f5d49a901c875b2e0114 (patch) | |
| tree | 83993038a7525f6b61aea8f76ea38ddb96325a2f /include/linux/libata.h | |
| parent | [libata] export sata_print_link_status() (diff) | |
| download | kernel-qcow2-linux-d88184fb2348a50f7c34f5d49a901c875b2e0114.tar.gz kernel-qcow2-linux-d88184fb2348a50f7c34f5d49a901c875b2e0114.tar.xz kernel-qcow2-linux-d88184fb2348a50f7c34f5d49a901c875b2e0114.zip | |
[libata] sata_mv: clean up DMA boundary issues, turn on 64-bit DMA
The chips covered by sata_mv have a 32-bit DMA boundary they must not
cross, not a 64K boundary. We are merely limited to a 64K maximum
segment size. Therefore, the DMA scatter/gather table fill code can be
greatly simplified, and we need not cut in half the S/G table size as
reported to the SCSI layer.
Also, the driver forget to turn on 64-bit DMA at the PCI layer. All
other data structures (both hardware and software) have been prepped for
64-bit PCI DMA. It was simply never turned on. <fingers crossed> let's
see if it still works...
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Diffstat (limited to 'include/linux/libata.h')
0 files changed, 0 insertions, 0 deletions