Hi Bjorn, On 2017/12/15 6:58, Bjorn Helgaas wrote:
Hi all, In the PCI config access path, the *_pcie_valid_device() functions in the dwc, altera, rockchip, and xilinx drivers all check whether the link is up. I think this is racy because the link may go down after we check but before we perform the config access.
The link could be broken at any time, so the check if bogus. And we have accessors for performing the config access but we don't have these for performing memory access, so again the racy is always there.
What would blow up if we removed the *_pcie_link_up() checks?
What Rockchip needs(probably for all arm64 system) is something like this patchset[1]. And for arm32, we could refer to imx6q_pcie_abort_handler in drivers/pci/dwc/pci-imx6.c Note that ATF(ARM trust firmware) could help capture these (a)synchronous, so we actually don't even need to resort to patchset[1], and handle these abort in the firmware. [1]https://lkml.org/lkml/2017/11/10/236
I'd like to either remove the checks or add comments about why the
Remove it makes sense to me.
race is acceptable. If we've covered this before, I apologize. Adding a comment will keep me from pestering you about this again in the future. Bjorn
