You are not logged in.
Hello,
I want to try out the new memory management tech that supposedly gives a 10% performance buff, but Nvidia DKMS fails to build. Here is the make.log file.
Offline
That's pretty common, you have to wait for nvidia to update for newer kernels.
Offline
See mm/migrate.c: remove MIGRATE_PFN_LOCKED https://github.com/torvalds/linux/commi … 34116f8d0f
Edit:
Should be fixable by adding the following to uvm_migrate_pageable.c before the first use of MIGRATE_PFN_LOCKED
#ifndef MIGRATE_PFN_LOCKED
#define MIGRATE_PFN_LOCKED 0
#endif
Last edited by loqs (2021-11-18 18:54:21)
Offline
See mm/migrate.c: remove MIGRATE_PFN_LOCKED https://github.com/torvalds/linux/commi … 34116f8d0f
In accordance with the link, I'd suggest a patch like this:
--- a/kernel/nvidia-uvm/uvm_migrate_pageable.c 2022-01-08 21:51:32.672454683 +0100
+++ b/kernel/nvidia-uvm/uvm_migrate_pageable.c 2022-01-08 21:51:52.298644944 +0100
@@ -406,7 +406,7 @@
uvm_push_set_flag(&push, UVM_PUSH_FLAG_CE_NEXT_MEMBAR_NONE);
copying_gpu->parent->ce_hal->memset_8(&push, dst_address, 0, PAGE_SIZE);
- dst[i] = migrate_pfn(page_to_pfn(dst_page)) | MIGRATE_PFN_LOCKED;
+ dst[i] = migrate_pfn(page_to_pfn(dst_page));
}
if (copying_gpu) {
@@ -490,7 +490,7 @@
uvm_push_set_flag(&push, UVM_PUSH_FLAG_CE_NEXT_MEMBAR_NONE);
copying_gpu->parent->ce_hal->memcopy(&push, dst_address, src_address, PAGE_SIZE);
- dst[i] = migrate_pfn(page_to_pfn(dst_page)) | MIGRATE_PFN_LOCKED;
+ dst[i] = migrate_pfn(page_to_pfn(dst_page));
}
// TODO: Bug 1766424: If the destination is a GPU and the copy was done by
Offline
Offline