On Mon, Jun 03, 2019 at 02:17:36PM -0700, JD Zheng wrote: > > On Sun, 2 Jun 2019 19:44:14 -0500 > > Bjorn Helgaas <helgaas@xxxxxxxxxx> wrote: > > > Would you mind opening a report at https://bugzilla.kernel.org and > > > attaching the complete dmesg log and "lspci -vv" output? > > I submitted one as https://bugzilla.kernel.org/show_bug.cgi?id=203797 > > > > Out of curiosity, why do you use "pciehp.pciehp_poll_time=5"? > > This was chosen by other engineer. I tried shorter polling time, which > doesn't make difference. > > I also tried pciehp.pciehp_poll_mode=0 but hotplug doesn't seem working. > Should I use poll mode or not? This is just a tangent to the real issue you're working; I'm not suggesting any changes here to fix that issue. In my mind, the existence of the pciehp.pciehp_poll_mode and pciehp.pciehp_poll_time parameters is a defect in pciehp. I would far rather that pciehp figured out by itself whether it needed to poll. I don't know whether that's actually feasible, or if there's some reason why pciehp can't be that smart. Bjorn
