On 02/02/2016 09:48 PM, Gavin Shan wrote:
On Tue, Jan 19, 2016 at 06:18:20PM -0200, Guilherme G. Piccoli wrote:
- /* only attempt to use a new window if 64-bit DMA is requested */
- if (!disable_ddw && dma_mask == DMA_BIT_MASK(64)) {
+ /* We should check if EEH is enabled here, since DDW mechanism has
+ * an intrinsic dependency of EEH config addr information. Also, we
+ * only attempt to use a new window if 64-bit DMA is requested */
+ if (eeh_enabled() && !disable_ddw && dma_mask == DMA_BIT_MASK(64)) {
dn = pci_device_to_OF_node(pdev);
dev_dbg(dev, "node is %s\n", dn->full_name);
There are two types of addresses: (1) PCI config address (2) PE config address.
(1) is used to indentify one PCI device which is included in the PE. (2) is the
PCI config address of PE's primary bus in pHyp. Both of them can be used to
identify
the PE. It means the (1) PCI config address, which is retrieved from pci_dn,
can be
passed to hypervisor. Then we don't have to disable DDW when EEH is disabled.
Guilherme, did you hit the crash on pHyp or PowerKVM?
Gavin, thanks very much for the clarification. So, we can interchange
edev->config_addr with pdn->pci_ext_config_space ?
This would solve the issue with DDW being enabled when EEH is not.
I hit the issue on PowerVM (PHyp). I wasn't able to perform hotplug in
qemu that time I was testing this - I can re-test on qemu. Can we use
pci_dn config address in qemu guest too?
Cheers,
Guilherme
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
