| Message ID | 20240529111904.2069608-1-david@redhat.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | mm: page_type, zsmalloc and page_mapcount_reset() | expand |
On (24/05/29 13:18), David Hildenbrand wrote: > Wanting to remove the remaining abuser of _mapcount/page_type along with > page_mapcount_reset(), I stumbled over zsmalloc, which is yet to be > converted away from "struct page" [1]. > > Unfortunately, we cannot stop using the page_type field in zsmalloc code > completely for its own purposes. All other fields in "struct page" are > used one way or the other. Could we simply store a 2-byte offset value > at the beginning of each page? Likely, but that will require a bit more > work; and once we have memdesc we might want to move the offset in there > (struct zsalloc?) again. > > ... but we can limit the abuse to 16 bit, glue it to a page type that > must be set, and document it. page_has_type() will always successfully > indicate such zsmalloc pages, and such zsmalloc pages only. > > We lose zsmalloc support for PAGE_SIZE > 64KB, which should be tolerable. > We could use more bits from the page type, but 16 bit sounds like a good > idea for now. > > So clarify the _mapcount/page_type documentation, use a proper page_type > for zsmalloc, and remove page_mapcount_reset(). > > Briefly tested with zram on x86-64. > > [1] https://lore.kernel.org/all/20231130101242.2590384-1-42.hyeyoo@gmail.com/ > > Cc: Andrew Morton <akpm@linux-foundation.org> > Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org> > Cc: Mike Rapoport <rppt@kernel.org> > Cc: Minchan Kim <minchan@kernel.org> > Cc: Sergey Senozhatsky <senozhatsky@chromium.org> > Cc: Hyeonggon Yoo <42.hyeyoo@gmail.com> Tested-by: Sergey Senozhatsky <senozhatsky@chromium.org> # zram/zsmalloc workloads
Wanting to remove the remaining abuser of _mapcount/page_type along with page_mapcount_reset(), I stumbled over zsmalloc, which is yet to be converted away from "struct page" [1]. Unfortunately, we cannot stop using the page_type field in zsmalloc code completely for its own purposes. All other fields in "struct page" are used one way or the other. Could we simply store a 2-byte offset value at the beginning of each page? Likely, but that will require a bit more work; and once we have memdesc we might want to move the offset in there (struct zsalloc?) again. ... but we can limit the abuse to 16 bit, glue it to a page type that must be set, and document it. page_has_type() will always successfully indicate such zsmalloc pages, and such zsmalloc pages only. We lose zsmalloc support for PAGE_SIZE > 64KB, which should be tolerable. We could use more bits from the page type, but 16 bit sounds like a good idea for now. So clarify the _mapcount/page_type documentation, use a proper page_type for zsmalloc, and remove page_mapcount_reset(). Briefly tested with zram on x86-64. [1] https://lore.kernel.org/all/20231130101242.2590384-1-42.hyeyoo@gmail.com/ Cc: Andrew Morton <akpm@linux-foundation.org> Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org> Cc: Mike Rapoport <rppt@kernel.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Hyeonggon Yoo <42.hyeyoo@gmail.com> v1 -> v2: * Rebased to mm/mm-unstable * "mm: update _mapcount and page_type documentation" -> Minor comment change * "mm: allow reuse of the lower 16 bit of the page type with an actual type" -> Fixup 18 vs 16 in description -> Reduce PAGE_TYPE_BASE to a single bit and hand-out bits from highest to lowest -> Adjust description RFC -> v1: * Rebased to v6.10-rc1 * "mm: update _mapcount and page_type documentation" -> Exchange members and fixup doc as suggested by Mike * "mm: allow reuse of the lower 16bit of the page type with an actual type" -> Remove "highest bit" comment, fixup PG_buddy, extend description * "mm/zsmalloc: use a proper page type" -> Add and use HAVE_ZSMALLOC to fixup compilcation -> Fixup BUILD_BUG_ON -> Add some VM_WARN_ON_ONCE(!PageZsmalloc(page)); * "mm/mm_init: initialize page->_mapcount directly in __init_single_page()" -> Fixup patch subject David Hildenbrand (6): mm: update _mapcount and page_type documentation mm: allow reuse of the lower 16 bit of the page type with an actual type mm/zsmalloc: use a proper page type mm/page_alloc: clear PageBuddy using __ClearPageBuddy() for bad pages mm/filemap: reinitialize folio->_mapcount directly mm/mm_init: initialize page->_mapcount directly in __init_single_page() drivers/block/zram/Kconfig | 1 + include/linux/mm.h | 10 ---------- include/linux/mm_types.h | 33 ++++++++++++++++++++++----------- include/linux/page-flags.h | 25 ++++++++++++++++--------- mm/Kconfig | 10 ++++++++-- mm/filemap.c | 2 +- mm/mm_init.c | 2 +- mm/page_alloc.c | 6 ++++-- mm/zsmalloc.c | 29 +++++++++++++++++++++++++---- 9 files changed, 78 insertions(+), 40 deletions(-)