| Message ID | 20220409073846.22286-1-linmiaohe@huawei.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | A few cleanup and fixup patches for migration | expand |
On Sat, 2022-04-09 at 15:38 +0800, Miaohe Lin wrote: > Hi everyone, > This series contains a few patches to remove unneeded lock page and > PageMovable check, reduce the rcu lock duration. Also we fix potential > pte_unmap on an not mapped pte. More details can be found in the > respective changelogs. Thanks! It appears that you ignored my comments for the previous version. Can you check it? Best Regards, Huang, Ying > --- > v1: > rebase [1] on mainline. > > [1] https://lore.kernel.org/lkml/20220304093409.25829-2-linmiaohe@huawei.com/T/ > --- > Miaohe Lin (4): > mm/migration: reduce the rcu lock duration > mm/migration: remove unneeded lock page and PageMovable check > mm/migration: return errno when isolate_huge_page failed > mm/migration: fix potential pte_unmap on an not mapped pte > > include/linux/migrate.h | 2 +- > include/linux/swapops.h | 4 ++-- > mm/filemap.c | 10 +++++----- > mm/hugetlb.c | 2 +- > mm/migrate.c | 31 +++++++++++++------------------ > 5 files changed, 22 insertions(+), 27 deletions(-) >
On 2022/4/12 10:25, ying.huang@intel.com wrote: > On Sat, 2022-04-09 at 15:38 +0800, Miaohe Lin wrote: >> Hi everyone, >> This series contains a few patches to remove unneeded lock page and >> PageMovable check, reduce the rcu lock duration. Also we fix potential >> pte_unmap on an not mapped pte. More details can be found in the >> respective changelogs. Thanks! > > It appears that you ignored my comments for the previous version. Can > you check it? I do remember [1] and I tried to make isolate_huge_page consistent with isolate_lru_page. But their return value conventions are different. isolate_huge_page return 0 when success while isolate_huge_page returns true in this case. So make them consistent would lead to many code change. I should have added this in my changelog. Thanks. [1] https://lore.kernel.org/linux-mm/8735jsgctq.fsf@yhuang6-desk2.ccr.corp.intel.com/ > > Best Regards, > Huang, Ying > >> --- >> v1: >> rebase [1] on mainline. >> >> [1] https://lore.kernel.org/lkml/20220304093409.25829-2-linmiaohe@huawei.com/T/ >> --- >> Miaohe Lin (4): >> mm/migration: reduce the rcu lock duration >> mm/migration: remove unneeded lock page and PageMovable check >> mm/migration: return errno when isolate_huge_page failed >> mm/migration: fix potential pte_unmap on an not mapped pte >> >> include/linux/migrate.h | 2 +- >> include/linux/swapops.h | 4 ++-- >> mm/filemap.c | 10 +++++----- >> mm/hugetlb.c | 2 +- >> mm/migrate.c | 31 +++++++++++++------------------ >> 5 files changed, 22 insertions(+), 27 deletions(-) >> > > > > . >
On Tue, 2022-04-12 at 11:29 +0800, Miaohe Lin wrote: > On 2022/4/12 10:25, ying.huang@intel.com wrote: > > On Sat, 2022-04-09 at 15:38 +0800, Miaohe Lin wrote: > > > Hi everyone, > > > This series contains a few patches to remove unneeded lock page and > > > PageMovable check, reduce the rcu lock duration. Also we fix potential > > > pte_unmap on an not mapped pte. More details can be found in the > > > respective changelogs. Thanks! > > > > It appears that you ignored my comments for the previous version. Can > > you check it? > > I do remember [1] and I tried to make isolate_huge_page consistent with isolate_lru_page. > But their return value conventions are different. isolate_huge_page return 0 when > success while isolate_huge_page returns true in this case. So make them consistent > would lead to many code change. I should have added this in my changelog. If you found new problem, you can reply to the original email. Just don't ignore the comments. Best Regards, Huang, Ying > Thanks. > > [1] https://lore.kernel.org/linux-mm/8735jsgctq.fsf@yhuang6-desk2.ccr.corp.intel.com/ > > > > > > Best Regards, > > Huang, Ying > > > > > --- > > > v1: > > > rebase [1] on mainline. > > > > > > [1] https://lore.kernel.org/lkml/20220304093409.25829-2-linmiaohe@huawei.com/T/ > > > --- > > > Miaohe Lin (4): > > > mm/migration: reduce the rcu lock duration > > > mm/migration: remove unneeded lock page and PageMovable check > > > mm/migration: return errno when isolate_huge_page failed > > > mm/migration: fix potential pte_unmap on an not mapped pte > > > > > > include/linux/migrate.h | 2 +- > > > include/linux/swapops.h | 4 ++-- > > > mm/filemap.c | 10 +++++----- > > > mm/hugetlb.c | 2 +- > > > mm/migrate.c | 31 +++++++++++++------------------ > > > 5 files changed, 22 insertions(+), 27 deletions(-) > > > > > > > > > > > . > > >
On Tue, 2022-04-12 at 11:29 +0800, Miaohe Lin wrote: > On 2022/4/12 10:25, ying.huang@intel.com wrote: > > On Sat, 2022-04-09 at 15:38 +0800, Miaohe Lin wrote: > > > Hi everyone, > > > This series contains a few patches to remove unneeded lock page and > > > PageMovable check, reduce the rcu lock duration. Also we fix potential > > > pte_unmap on an not mapped pte. More details can be found in the > > > respective changelogs. Thanks! > > > > It appears that you ignored my comments for the previous version. Can > > you check it? > > I do remember [1] and I tried to make isolate_huge_page consistent with isolate_lru_page. > But their return value conventions are different. isolate_huge_page return 0 when > success while isolate_huge_page returns true in this case. So make them consistent > would lead to many code change. I should have added this in my changelog. I found that there are only 7 callers of isolate_huge_page(). It sounds like something that is still doable. Best Regards, Huang, Ying > Thanks. > > [1] https://lore.kernel.org/linux-mm/8735jsgctq.fsf@yhuang6-desk2.ccr.corp.intel.com/ > > > > > > Best Regards, > > Huang, Ying > > > > > --- > > > v1: > > > rebase [1] on mainline. > > > > > > [1] https://lore.kernel.org/lkml/20220304093409.25829-2-linmiaohe@huawei.com/T/ > > > --- > > > Miaohe Lin (4): > > > mm/migration: reduce the rcu lock duration > > > mm/migration: remove unneeded lock page and PageMovable check > > > mm/migration: return errno when isolate_huge_page failed > > > mm/migration: fix potential pte_unmap on an not mapped pte > > > > > > include/linux/migrate.h | 2 +- > > > include/linux/swapops.h | 4 ++-- > > > mm/filemap.c | 10 +++++----- > > > mm/hugetlb.c | 2 +- > > > mm/migrate.c | 31 +++++++++++++------------------ > > > 5 files changed, 22 insertions(+), 27 deletions(-) > > > > > > > > > > > . > > >
On 2022/4/12 14:33, ying.huang@intel.com wrote: > On Tue, 2022-04-12 at 11:29 +0800, Miaohe Lin wrote: >> On 2022/4/12 10:25, ying.huang@intel.com wrote: >>> On Sat, 2022-04-09 at 15:38 +0800, Miaohe Lin wrote: >>>> Hi everyone, >>>> This series contains a few patches to remove unneeded lock page and >>>> PageMovable check, reduce the rcu lock duration. Also we fix potential >>>> pte_unmap on an not mapped pte. More details can be found in the >>>> respective changelogs. Thanks! >>> >>> It appears that you ignored my comments for the previous version. Can >>> you check it? >> >> I do remember [1] and I tried to make isolate_huge_page consistent with isolate_lru_page. >> But their return value conventions are different. isolate_huge_page return 0 when >> success while isolate_huge_page returns true in this case. So make them consistent >> would lead to many code change. I should have added this in my changelog. > > If you found new problem, you can reply to the original email. Just > don't ignore the comments. Sorry about it. I would take care of this next time. Thanks! > > Best Regards, > Huang, Ying > >> Thanks. >> >> [1] https://lore.kernel.org/linux-mm/8735jsgctq.fsf@yhuang6-desk2.ccr.corp.intel.com/ >> >> >>> >>> Best Regards, >>> Huang, Ying >>> >>>> --- >>>> v1: >>>> rebase [1] on mainline. >>>> >>>> [1] https://lore.kernel.org/lkml/20220304093409.25829-2-linmiaohe@huawei.com/T/ >>>> --- >>>> Miaohe Lin (4): >>>> mm/migration: reduce the rcu lock duration >>>> mm/migration: remove unneeded lock page and PageMovable check >>>> mm/migration: return errno when isolate_huge_page failed >>>> mm/migration: fix potential pte_unmap on an not mapped pte >>>> >>>> include/linux/migrate.h | 2 +- >>>> include/linux/swapops.h | 4 ++-- >>>> mm/filemap.c | 10 +++++----- >>>> mm/hugetlb.c | 2 +- >>>> mm/migrate.c | 31 +++++++++++++------------------ >>>> 5 files changed, 22 insertions(+), 27 deletions(-) >>>> >>> >>> >>> >>> . >>> >> > > > . >
On 2022/4/12 15:00, ying.huang@intel.com wrote: > On Tue, 2022-04-12 at 11:29 +0800, Miaohe Lin wrote: >> On 2022/4/12 10:25, ying.huang@intel.com wrote: >>> On Sat, 2022-04-09 at 15:38 +0800, Miaohe Lin wrote: >>>> Hi everyone, >>>> This series contains a few patches to remove unneeded lock page and >>>> PageMovable check, reduce the rcu lock duration. Also we fix potential >>>> pte_unmap on an not mapped pte. More details can be found in the >>>> respective changelogs. Thanks! >>> >>> It appears that you ignored my comments for the previous version. Can >>> you check it? >> >> I do remember [1] and I tried to make isolate_huge_page consistent with isolate_lru_page. >> But their return value conventions are different. isolate_huge_page return 0 when >> success while isolate_huge_page returns true in this case. So make them consistent >> would lead to many code change. I should have added this in my changelog. > > I found that there are only 7 callers of isolate_huge_page(). It sounds > like something that is still doable. Fine, I will try to do this. :) Thanks! > > Best Regards, > Huang, Ying > >> Thanks. >> >> [1] https://lore.kernel.org/linux-mm/8735jsgctq.fsf@yhuang6-desk2.ccr.corp.intel.com/ >> >> >>> >>> Best Regards, >>> Huang, Ying >>> >>>> --- >>>> v1: >>>> rebase [1] on mainline. >>>> >>>> [1] https://lore.kernel.org/lkml/20220304093409.25829-2-linmiaohe@huawei.com/T/ >>>> --- >>>> Miaohe Lin (4): >>>> mm/migration: reduce the rcu lock duration >>>> mm/migration: remove unneeded lock page and PageMovable check >>>> mm/migration: return errno when isolate_huge_page failed >>>> mm/migration: fix potential pte_unmap on an not mapped pte >>>> >>>> include/linux/migrate.h | 2 +- >>>> include/linux/swapops.h | 4 ++-- >>>> mm/filemap.c | 10 +++++----- >>>> mm/hugetlb.c | 2 +- >>>> mm/migrate.c | 31 +++++++++++++------------------ >>>> 5 files changed, 22 insertions(+), 27 deletions(-) >>>> >>> >>> >>> >>> . >>> >> > > > > . >