On Thu, May 03, 2018 at 15:37:34 +0200, Christian König wrote: > Am 26.04.2018 um 08:45 schrieb Sami Farin: > > On Wed, Apr 25, 2018 at 16:59:43 -0400, Alex Deucher wrote: > > > On Wed, Apr 25, 2018 at 1:13 PM, Sami Farin <hvtaifwkbgefbaei at gmail.com> wrote: > > > > Is it supposed to work? Now if I use that option, monitor powers off > > > > (no signal) when it's initializing amdgpu and I have to reset the computer. > > > > Instead of crashing, could it not just give an error or something? > > > Do you have the IOMMU enabled? I'm guessing there is some issue in > > > the IOMMU code or in TTM causing the C bit (bit 47) to get dropped > > > somewhere. Your GPU only supports a 40 bit DMA mask, so either IOMMU > > > or bounce buffers would be required to remap the address into a range > > > the GPU could access. > > > > > > Alex > > I have IOMMU. > > Mhm, from those logs at least in theory it should work fine. > > There really isn't any other error message and the system isn't accessible > using SSH over the network? Well, sysrq does not work, I have to press reset button, so I presume ssh would not work, either. > I would guess that we get at least some kind of error message or failure to > start the driver, but a complete crash is rather unexpected. Well I could try boot_delay kernel param, or if you have suspicion about some function I could add some printk + delay there. > Regards, > Christian. -- Do what you love because life is too short for anything else.
