On 3/31/2008, "Dave Liu" r63238@freescale.com wrote:
Don't do this.
I don't think the ATA_SECT_BYTESIZE is better than the ATA_SECT_SIZE.
The big problem here is that ata.h defines ATA_SECT_SIZE as 256 Words where libata defines it as 512 Bytes.
I did not want to change the old define since it would possibly break more than the fsl driver.
When I grep 'ATA_SECT_SIZE', I can *not* find any files including the ATA_SECT_SIZE before the fsl_sata.c.
Well, the ata_piix driver for one uses it, but only includes ata.h. But that is the only one in the tree right now.
But the problem still remains, it is currently impossible to use both libata.h and ata.h.
Why need use both them? I believe the libata.h is enough.
Since my driver for the sil3114 uses "legacy mode" it uses a lot of the ata commands directly in PIO mode. These are defined in ata.h (Both in u-boot and Linux)
If the libata.h is not enough for you, I suggest you update it from the latest linux kernel.
Still, the kernel has both a ata.h and a libata.h, where libata.h includes ata.h. And where ATA_SECT_SIZE, and the other defines i use, actually is defined in ata.h albeit as 512 bytes :).
I think it is correct way to handle it. It will be more easy to to maintain the header.
I have no problem with either solution. If its ok to add the defines to libata.h instead i will do that?
/Tor