On 03/12/19, 6:28 PM, "Vlastimil Babka" <vbabka@xxxxxxx> wrote: >>> >>> [ 4.4 backport: there's get_page_foll(), so add try_get_page()-like checks >>> in there, enabled by a new parameter, which is false where >>> upstream patch doesn't replace get_page() with try_get_page() >>> (the THP and hugetlb callers). >> >> Could we have try_get_page_foll(), as in: >> https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.kernel.org%2Fstable%2F1570581863-12090-3-git-send-email-akaher%40vmware.com%2F&data=02%7C01%7Cakaher%40vmware.com%7Cb6592f0fbec040aa045f08d777f06a9f%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C0%7C0%7C637109746821395444&sdata=cYBj3SvEikPbiHsVZj3zCys8t9ISLiHKzAlsSqiZRW8%3D&reserved=0 >> >> + Code will be in sync as we have try_get_page() >> + No need to add extra argument to try_get_page() >> + No need to modify the callers of try_get_page() Any reason for not using try_get_page_foll(). >>> In gup_pte_range(), we don't expect tail pages, so just check >>> page ref count instead of try_get_compound_head() >> >> Technically it's fine. If you want to keep the code of stable versions in sync >> with latest versions then this could be done in following ways (without any >> modification in upstream patch for gup_pte_range()): >> >> Apply 7aef4172c7957d7e65fc172be4c99becaef855d4 before applying >> 8fde12ca79aff9b5ba951fce1a2641901b8d8e64, as done here: >> https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.kernel.org%2Fstable%2F1570581863-12090-4-git-send-email-akaher%40vmware.com%2F&data=02%7C01%7Cakaher%40vmware.com%7Cb6592f0fbec040aa045f08d777f06a9f%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C0%7C0%7C637109746821395444&sdata=gTJMJ3Yx6G0ng46TQsBzCS2DowwP7YtIjluKJuqvN6o%3D&reserved=0 > Yup, I have considered that, and deliberately didn't add that commit > 7aef4172c795 ("mm: handle PTE-mapped tail pages in gerneric fast gup > implementaiton") as it's part of a large THP refcount rework. In 4.4 we > don't expect to GUP tail pages so I wanted to keep it that way - > minimally, the compound_head() operation is a unnecessary added cost, > although it would also work.
