Jesse Barnes wrote:
Back in July, Adrian Bunk sent in a patch to make SCSI_SATA tristate. This
prevents the intel_ide_combined quirk in drivers/pci/quirks.c from working if
SCSI_SATA=m, which is the case for Fedora kernels (my motivation for tracking
this down).
In my configuration, not running the quirk causes the ata_piix driver (the
libata driver for my IDE controller) to fail to attach to the device, since
the legacy IDE driver has already claimed the ports. Unfortunately, the AHCI
driver also tries to mess with the device, and ends up disabling its
interrupts before aborting its load, causing the IDE layer to complain loudly
that hda is losing interrupts.
Thanks a ton for figuring this out!
So what should be done? Ideally, libata would fully support ATAPI and then I
wouldn't need the legacy IDE drivers at all on this box, making the quirk
moot, but that won't happen for 2.6.14, so we'll need something else.
Unconditionally enabling the quirk will cause at least one of the ports to be
reserved for the SATA driver, which may never load. And obviously not
running the quirk leads to the situation described above.
A hack that might be suitable for 2.6.14 is to make the quirk depend on either
CONFIG_SCSI_SATA or CONFIG_SCSI_SATA_MODULE. Then the quirk could be removed
entirely when ATAPI support for libata is merged.
Overall the quirk is a hack until ATAPI is supported -- but even after
ATAPI is supported, we need to figure out some way to keep IDE driver
from stealing the legacy IDE ports before libata can touch them :(
However, when ATAPI is supported, that at least means that both PATA and
SATA can run at full speed.
Your patch is correct, and should go into 2.6.14-rc post-haste.
I continue to grumble at the Kconfig annoyance: CONFIG_SCSI_SATA is
fundamentally a yes/no question. CONFIG_SCSI_SATA=m makes no sense at
all -- and causes problems, as we see -- but is required in order to do
Kconfig dependencies correctly AFAIK.
Maybe 'if SCSI_SATA' is needed instead? A lame way to implement
dependencies, but if there is no other way...
Jeff
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[Index of Archives]
[Kernel Newbies]
[Netfilter]
[Bugtraq]
[Photo]
[Stuff]
[Gimp]
[Yosemite News]
[MIPS Linux]
[ARM Linux]
[Linux Security]
[Linux RAID]
[Video 4 Linux]
[Linux for the blind]
[Linux Resources]