On Tue, 2025-08-12 at 11:17 +0200, Christian König wrote: > On 11.08.25 17:16, Thomas Hellström wrote: > > > > > > > FWIW If I understand the code correctly, i915 bypasses this by > > > > setting > > > > up user-space mappings not by vm_insert_pfn_prot() but using > > > > apply_to_page_range(), mapping the whole bo. > > > > > > Yeah, that's probably not something we can do. Even filling in > > > 2MiB > > > of address space at a time caused performance problems for TTM. > > > > Wasn't that because of repeated calls to vmf_insert_pfn_prot(), > > repeating the caching checks and page-table walk all the time? > > Only partially, the main problem was that only a fraction of the BO > was actually CPU accessed. So filling in more than faulted was just > overhead. > > > I think apply_to_page_range() should be pretty fast. Also, to avoid > > regressions due to changing the number of prefaulted pages, we > > could > > perhaps honor the MADV_RANDOM and MADV_SEQUENTIAL advises for UMD > > to > > use; one faulting a single page only, one faulting the whole bo > > Ah! In my thinking apply_to_page_range() would always fault in the > whole BO, if that still works for only a partial range than that > should be ok.
Yes, it looks like it works with partial ranges. > > > , but > > also see below: > > > > > > > > We should probably just drop overriding the attributes in > > > vmf_insert_pfn_prot(). > > > > I think either solution will see resistance with arch people. We > > should > > probably involve them in the discussion. > > Any specific person I should CC or just x...@kernel.org? scripts/get_maintainer.pl of a tiny change into the PAT code gives me the following: Hopefully some of these should trigger some replies and insights: ./scripts/get_maintainer.pl 0001-arch-mm-dummy-commit.patch Dave Hansen <dave.han...@linux.intel.com> (maintainer:X86 MM) Andy Lutomirski <l...@kernel.org> (maintainer:X86 MM) Peter Zijlstra <pet...@infradead.org> (maintainer:X86 MM) Thomas Gleixner <t...@linutronix.de> (maintainer:X86 ARCHITECTURE (32- BIT AND 64-BIT)) Ingo Molnar <mi...@redhat.com> (maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT),commit_signer:10/16=62%,authored:2/16=12%) Borislav Petkov <b...@alien8.de> (maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT)) x...@kernel.org (maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT)) "H. Peter Anvin" <h...@zytor.com> (reviewer:X86 ARCHITECTURE (32-BIT AND 64-BIT)) Andrew Morton <a...@linux-foundation.org> (commit_signer:7/16=44%) David Hildenbrand <da...@redhat.com> (commit_signer:6/16=38%,authored:6/16=38%,added_lines:68/105=65%,remove d_lines:214/292=73%) "Liam R. Howlett" <liam.howl...@oracle.com> (commit_signer:5/16=31%) Lorenzo Stoakes <lorenzo.stoa...@oracle.com> (commit_signer:5/16=31%) Sean Christopherson <sea...@google.com> (authored:2/16=12%,added_lines:15/105=14%,removed_lines:31/292=11%) "Thomas Hellström" <thomas.hellst...@linux.intel.com> (authored:1/16=6%) Shivank Garg <shiva...@amd.com> (authored:1/16=6%) Peter Xu <pet...@redhat.com> (added_lines:7/105=7%) Dan Williams <dan.j.willi...@intel.com> (removed_lines:27/292=9%) > > Thanks, > Christian
