On 3/3/22 11:09, Dave Hansen wrote:
On 2/24/22 08:56, Brijesh Singh wrote:+ /* + * Allocate VMSA page to work around the SNP erratum where the CPU will + * incorrectly signal an RMP violation #PF if a large page (2MB or 1GB) + * collides with the RMP entry of VMSA page. The recommended workaround + * is to not use a large page. + * + * Allocate one extra page, use a page which is not 2MB-aligned + * and free the other. + */ + p = alloc_pages(GFP_KERNEL_ACCOUNT | __GFP_ZERO, 1); + if (!p) + return NULL; + + split_page(p, 1); + + pfn = page_to_pfn(p); + if (IS_ALIGNED(__pfn_to_phys(pfn), PMD_SIZE)) { + pfn++; + __free_page(p); + } else { + __free_page(pfn_to_page(pfn + 1)); + } + + return page_address(pfn_to_page(pfn)); +}This can be simplified. There's no need for all the sill pfn_to_page() conversions or even an alignment check. The second page (page[1]) of an order-1 page is never 2M/1G aligned. Just use that: // Alloc an 8k page which is also 8k-aligned: p = alloc_pages(GFP_KERNEL_ACCOUNT | __GFP_ZERO, 1); if (!p) return NULL; split_page(p, 1); // Free the first 4k. This page _may_ // be 2M/1G aligned and can not be used: __free_page(p); // Return the unaligned page: return page_address(p+1);
Ack, I am good with using your version. -Brijesh
