On Tue, Feb 04, 2025 at 09:41:29PM +0100, David Hildenbrand wrote: > On 04.02.25 21:26, Jason Gunthorpe wrote: > > On Tue, Feb 04, 2025 at 09:05:47PM +0100, David Hildenbrand wrote: > > > Fully agreed, this is going into the right direction. Dumping what's mapped > > > is a different story. Maybe that dumping logic could simply be written in C > > > for the time being? > > > > ? > > > > Isn't dumping just a > > decode pte -> phys_to_virt() -> for_each_u64(virt) -> printk? > > > > IIUC, the problematic bit is that you might not have a directmap such that > phys_to_virt() would tell you the whole story. The phys_to_virt() I mean is on the page table pages themselves, not on a leaf. All page table pages came from alloc_pages_node() and you'd do virt_to_phys() to stick them into a next-table PTE, then phys_to_virt() to bring them back: phys_to_virt(virt_to_phys(page_address(alloc_pages_node()))) Effectively. The leaf pages could be anything and you should enver phys_to_virt those, this is what I mean by: > > Or do you mean trying to decode the leaf entires into some struct page > > detail? Does a GPU need to do that? Agree that would be better as some > > C function to take in a KVA instead of a struct page and populate a > > string with details about that KVA from the struct page. Though I should have said phys_addr_t not KVA Jason
