On Thu, Apr 22, 2010 at 09:23:10PM -0600, Robert Hancock wrote:
> Hmm, that seems like this line in drivers/ata/sata_sil24.c:
>
> /* Set max read request size to 4096. This slightly increases
> * write throughput for pci-e variants.
> */
> pcie_set_readrq(pdev, 4096);
>
> Can you try commenting out or deleting that line and rebuilding, and see
> if things work better? That's not really a solution, but if it worked,
> it would indicate there's something about the system configuration with
> that converter card installed that explodes when trying to set the max
> read request size. In that case can you post the output of "lspci -vv"
> and the full dmesg boot log?
Having put printk messages all over the driver, I see it currently locks up
during this readl call:
/* Clear port RST */
tmp = readl(port + PORT_CTRL_STAT);
.. but as this is clearly later on from the previously mentioned stack, I'm
starting to suspect that this is failing in random places when talking across
this pci-pcie bridge. Do you agree that this sounds plausible?
I did attempt to get a crash dump, but failed to get any joy from using the
nmi watchdog to trigger the use of the debug kernel in this state.
I will attempt using a different machine, hopefully tomorrow, but I have no
native pcie machines so can't eliminate the pci-pcie bridge being at fault.
Thanks for your help.
--
Richard Mawson
Mawson IT Services
Attachment:
signature.asc
Description: Digital signature
