Closed kennystrawnmusic closed 2 years ago
th0ma7
commented
5 years ago Just faced the same issue, not with claymore specifically but rather with amdgpu 18.50 drivers. 18.40 worked all right but launching clinfo after 18.50 upgrade creates a segmentation fault. Any work around from your end on this?
sstativa
commented
5 years ago I had the same problem with 18.50 in Ubuntu 18.04 and I managed to solve it in some way. However, it was before Christmas and I don't remember details.
There is only one command left in my notes, I believe it was a part of the solution. Try to install the driver according to the official guide then install this manually
sudo dpkg -i amdgpu-core_18.50-708488_all.deb \
libdrm2-amdgpu_2.4.95-708488_amd64.deb \
libdrm-amdgpu-common_1.0.0-708488_all.deb \
libdrm-amdgpu-amdgpu1_2.4.95-708488_amd64.deb \
libdrm-amdgpu-dev_2.4.95-708488_amd64.deb \
libdrm-amdgpu-radeon1_2.4.95-708488_amd64.deb
In a previous installation, before I screwed it up and had to start all over, I was running Claymore under Kubuntu 18.04 and the AMDGPU-Pro 18.30 driver just fine. Now that I have the new system and the only version of the driver I'm able to download is 18.50, I'm running into this problem:
Upon further inspection (via gdb), what's causing the segfault appears to be a null pointer dereference ― either in the miner or, most likely given that I'm still using the same installation both before and after this system reinstallation, in one of the OpenCL components of the 18.50 driver itself:
As you can see, something in there, either in Claymore or in one of the dependencies (again, system level dependencies most likely given that it's the same version), is trying to access memory at 0x0000000000000000 ― the address of a null pointer. If this is within the control of Claymore developers, then great, it's an easy fix. If not, and if it's a driver issue, then how can this information be forwarded over to AMD for them to take a look at it?