Message ID | 20241203102050.223318-1-david@redhat.com (mailing list archive) |
---|---|
State | New |
Headers | show |
Series | [RESEND,v1] mm/memory_hotplug: move debug_pagealloc_map_pages() into online_pages_range() | expand |
On Tue, Dec 03, 2024 at 11:20:50AM +0100, David Hildenbrand wrote: > In the near future, we want to have a single way to handover PageOffline > pages to the buddy, whereby they could have: > > (a) Never been exposed to the buddy before: kept PageOffline when onlining > the memory block. > (b) Been allocated from the buddy, for example using > alloc_contig_range() to then be set PageOffline, > > Let's start by making generic_online_page() less special compared to This got me confused for a moment. I'd have said "making __free_pages_core" less special? As that is where we were calling debug_pagealloc_map_pages() if the context on the context. > ordinary page freeing (e.g., free_contig_range()), and perform the > debug_pagealloc_map_pages() call unconditionally, even when the online > callback might decide to keep the pages offline. > > All pages are already initialized with PageOffline, so nobody touches > them either way. > > Cc: Andrew Morton <akpm@linux-foundation.org> > Cc: Oscar Salvador <osalvador@suse.de> > Signed-off-by: David Hildenbrand <david@redhat.com> Acked-by: Oscar Salvador <osalvador@suse.de>
diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c index c43b4e7fb298..20af14e695c7 100644 --- a/mm/memory_hotplug.c +++ b/mm/memory_hotplug.c @@ -650,6 +650,7 @@ static void online_pages_range(unsigned long start_pfn, unsigned long nr_pages) * this and the first chunk to online will be pageblock_nr_pages. */ for (pfn = start_pfn; pfn < end_pfn;) { + struct page *page = pfn_to_page(pfn); int order; /* @@ -664,7 +665,14 @@ static void online_pages_range(unsigned long start_pfn, unsigned long nr_pages) else order = MAX_PAGE_ORDER; - (*online_page_callback)(pfn_to_page(pfn), order); + /* + * Exposing the page to the buddy by freeing can cause + * issues with debug_pagealloc enabled: some archs don't + * like double-unmappings. So treat them like any pages that + * were allocated from the buddy. + */ + debug_pagealloc_map_pages(page, 1 << order); + (*online_page_callback)(page, order); pfn += (1UL << order); } diff --git a/mm/page_alloc.c b/mm/page_alloc.c index cc3296cf8c95..01927f03af0b 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -1293,12 +1293,6 @@ void __meminit __free_pages_core(struct page *page, unsigned int order, set_page_count(p, 0); } - /* - * Freeing the page with debug_pagealloc enabled will try to - * unmap it; some archs don't like double-unmappings, so - * map it first. - */ - debug_pagealloc_map_pages(page, nr_pages); adjust_managed_page_count(page, nr_pages); } else { for (loop = 0; loop < nr_pages; loop++, p++) {
In the near future, we want to have a single way to handover PageOffline pages to the buddy, whereby they could have: (a) Never been exposed to the buddy before: kept PageOffline when onlining the memory block. (b) Been allocated from the buddy, for example using alloc_contig_range() to then be set PageOffline, Let's start by making generic_online_page() less special compared to ordinary page freeing (e.g., free_contig_range()), and perform the debug_pagealloc_map_pages() call unconditionally, even when the online callback might decide to keep the pages offline. All pages are already initialized with PageOffline, so nobody touches them either way. Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Oscar Salvador <osalvador@suse.de> Signed-off-by: David Hildenbrand <david@redhat.com> --- SMTP server issues, resending so it reaches linux-mm as well. --- mm/memory_hotplug.c | 10 +++++++++- mm/page_alloc.c | 6 ------ 2 files changed, 9 insertions(+), 7 deletions(-)