| Message ID | 20220308033009.1400464-1-longman@redhat.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | mm/list_lru: Optimize memcg_drain_list_lru_node() | expand |
On Mon, Mar 07, 2022 at 10:30:09PM -0500, Waiman Long wrote: > Since commit 2c80cd57c743 ("mm/list_lru.c: fix list_lru_count_node() > to be race free"), we are tracking the total number of lru > entries in a list_lru_node in its nr_items field. In the case of > memcg_drain_list_lru_node(), there is nothing to be done if nr_items > is 0. We don't even need to take the nlru->lock as no new lru entry > could be added by a racing list_lru_add() to the draining src_idx memcg > at this point. > > Signed-off-by: Waiman Long <longman@redhat.com> Hi Waiman! The patch makes total sense to me, however it needs to be rebased at least on top of "mm: list_lru: rename memcg_drain_all_list_lrus to memcg_reparent_list_lrus". Thanks! > --- > mm/list_lru.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/mm/list_lru.c b/mm/list_lru.c > index 0cd5e89ca063..100ca453e885 100644 > --- a/mm/list_lru.c > +++ b/mm/list_lru.c > @@ -518,6 +518,12 @@ static void memcg_drain_list_lru_node(struct list_lru *lru, int nid, > int dst_idx = dst_memcg->kmemcg_id; > struct list_lru_one *src, *dst; > > + /* > + * If there is no lru entry in this nlru, we can skip it immediately. > + */ > + if (!READ_ONCE(nlru->nr_items)) > + return; > + > /* > * Since list_lru_{add,del} may be called under an IRQ-safe lock, > * we have to use IRQ-safe primitives here to avoid deadlock. > -- > 2.27.0 > >
On 3/7/22 23:39, Roman Gushchin wrote: > On Mon, Mar 07, 2022 at 10:30:09PM -0500, Waiman Long wrote: >> Since commit 2c80cd57c743 ("mm/list_lru.c: fix list_lru_count_node() >> to be race free"), we are tracking the total number of lru >> entries in a list_lru_node in its nr_items field. In the case of >> memcg_drain_list_lru_node(), there is nothing to be done if nr_items >> is 0. We don't even need to take the nlru->lock as no new lru entry >> could be added by a racing list_lru_add() to the draining src_idx memcg >> at this point. >> >> Signed-off-by: Waiman Long <longman@redhat.com> > Hi Waiman! > > The patch makes total sense to me, however it needs to be rebased at least > on top of "mm: list_lru: rename memcg_drain_all_list_lrus to memcg_reparent_list_lrus". > > Thanks! > > This patch was based on the current linux-next tree which includes commit ff221bc26bdd ("mm: list_lru: rename memcg_drain_all_list_lrus to memcg_reparent_list_lrus"). I do remember to double-check linux-next before sending this patch out. In fact, the same patch can be applied to both linux and linux-next tree without problem. Cheers, Longman
On Tue, Mar 08, 2022 at 10:41:38AM -0500, Waiman Long wrote: > On 3/7/22 23:39, Roman Gushchin wrote: > > On Mon, Mar 07, 2022 at 10:30:09PM -0500, Waiman Long wrote: > > > Since commit 2c80cd57c743 ("mm/list_lru.c: fix list_lru_count_node() > > > to be race free"), we are tracking the total number of lru > > > entries in a list_lru_node in its nr_items field. In the case of > > > memcg_drain_list_lru_node(), there is nothing to be done if nr_items > > > is 0. We don't even need to take the nlru->lock as no new lru entry > > > could be added by a racing list_lru_add() to the draining src_idx memcg > > > at this point. > > > > > > Signed-off-by: Waiman Long <longman@redhat.com> > > Hi Waiman! > > > > The patch makes total sense to me, however it needs to be rebased at least > > on top of "mm: list_lru: rename memcg_drain_all_list_lrus to memcg_reparent_list_lrus". > > > > Thanks! > > > > > This patch was based on the current linux-next tree which includes commit > ff221bc26bdd ("mm: list_lru: rename memcg_drain_all_list_lrus to > memcg_reparent_list_lrus"). I do remember to double-check linux-next before > sending this patch out. In fact, the same patch can be applied to both linux > and linux-next tree without problem. I'm looking at the mm tree (https://github.com/hnaz/linux-mm.git) and clearly see that commit "mm: list_lru: rename memcg_drain_all_list_lrus to memcg_reparent_list_lrus" eliminated the function with the name memcg_drain_list_lru_node(), which your patch is touching. Currently the function is located in list_lru.c and is named memcg_reparent_list_lru_node(). linux-next is sometimes a bit behind the mm tree. Thanks!
On 3/8/22 19:17, Roman Gushchin wrote: > On Tue, Mar 08, 2022 at 10:41:38AM -0500, Waiman Long wrote: >> On 3/7/22 23:39, Roman Gushchin wrote: >>> On Mon, Mar 07, 2022 at 10:30:09PM -0500, Waiman Long wrote: >>>> Since commit 2c80cd57c743 ("mm/list_lru.c: fix list_lru_count_node() >>>> to be race free"), we are tracking the total number of lru >>>> entries in a list_lru_node in its nr_items field. In the case of >>>> memcg_drain_list_lru_node(), there is nothing to be done if nr_items >>>> is 0. We don't even need to take the nlru->lock as no new lru entry >>>> could be added by a racing list_lru_add() to the draining src_idx memcg >>>> at this point. >>>> >>>> Signed-off-by: Waiman Long <longman@redhat.com> >>> Hi Waiman! >>> >>> The patch makes total sense to me, however it needs to be rebased at least >>> on top of "mm: list_lru: rename memcg_drain_all_list_lrus to memcg_reparent_list_lrus". >>> >>> Thanks! >>> >>> >> This patch was based on the current linux-next tree which includes commit >> ff221bc26bdd ("mm: list_lru: rename memcg_drain_all_list_lrus to >> memcg_reparent_list_lrus"). I do remember to double-check linux-next before >> sending this patch out. In fact, the same patch can be applied to both linux >> and linux-next tree without problem. > I'm looking at the mm tree (https://github.com/hnaz/linux-mm.git) and clearly > see that commit "mm: list_lru: rename memcg_drain_all_list_lrus to > memcg_reparent_list_lrus" eliminated the function with the name > memcg_drain_list_lru_node(), which your patch is touching. > Currently the function is located in list_lru.c and is named > memcg_reparent_list_lru_node(). > > linux-next is sometimes a bit behind the mm tree. Oh, you are right. I will rebase the patch based on linux-mm. Thanks for the suggestion. Cheers, Longman
On 3/8/22 19:50, Waiman Long wrote: > On 3/8/22 19:17, Roman Gushchin wrote: >> On Tue, Mar 08, 2022 at 10:41:38AM -0500, Waiman Long wrote: >>> On 3/7/22 23:39, Roman Gushchin wrote: >>>> On Mon, Mar 07, 2022 at 10:30:09PM -0500, Waiman Long wrote: >>>>> Since commit 2c80cd57c743 ("mm/list_lru.c: fix list_lru_count_node() >>>>> to be race free"), we are tracking the total number of lru >>>>> entries in a list_lru_node in its nr_items field. In the case of >>>>> memcg_drain_list_lru_node(), there is nothing to be done if nr_items >>>>> is 0. We don't even need to take the nlru->lock as no new lru entry >>>>> could be added by a racing list_lru_add() to the draining src_idx >>>>> memcg >>>>> at this point. >>>>> >>>>> Signed-off-by: Waiman Long <longman@redhat.com> >>>> Hi Waiman! >>>> >>>> The patch makes total sense to me, however it needs to be rebased >>>> at least >>>> on top of "mm: list_lru: rename memcg_drain_all_list_lrus to >>>> memcg_reparent_list_lrus". >>>> >>>> Thanks! >>>> >>>> >>> This patch was based on the current linux-next tree which includes >>> commit >>> ff221bc26bdd ("mm: list_lru: rename memcg_drain_all_list_lrus to >>> memcg_reparent_list_lrus"). I do remember to double-check linux-next >>> before >>> sending this patch out. In fact, the same patch can be applied to >>> both linux >>> and linux-next tree without problem. >> I'm looking at the mm tree (https://github.com/hnaz/linux-mm.git) and >> clearly >> see that commit "mm: list_lru: rename memcg_drain_all_list_lrus to >> memcg_reparent_list_lrus" eliminated the function with the name >> memcg_drain_list_lru_node(), which your patch is touching. >> Currently the function is located in list_lru.c and is named >> memcg_reparent_list_lru_node(). >> >> linux-next is sometimes a bit behind the mm tree. > > Oh, you are right. I will rebase the patch based on linux-mm. > > Thanks for the suggestion. The patch was originally based on the linux tree. Then I applied it onto linux-next without any issue. Unfortunately, I failed to notice that the hunk was applied to memcg_reparent_list_lru_node() without realizing that it is the new name for the old memcg_drain_list_lru_node(). The linux-next tree that I used does have all the necessary patches. Cheers, Longman
diff --git a/mm/list_lru.c b/mm/list_lru.c index 0cd5e89ca063..100ca453e885 100644 --- a/mm/list_lru.c +++ b/mm/list_lru.c @@ -518,6 +518,12 @@ static void memcg_drain_list_lru_node(struct list_lru *lru, int nid, int dst_idx = dst_memcg->kmemcg_id; struct list_lru_one *src, *dst; + /* + * If there is no lru entry in this nlru, we can skip it immediately. + */ + if (!READ_ONCE(nlru->nr_items)) + return; + /* * Since list_lru_{add,del} may be called under an IRQ-safe lock, * we have to use IRQ-safe primitives here to avoid deadlock.
Since commit 2c80cd57c743 ("mm/list_lru.c: fix list_lru_count_node() to be race free"), we are tracking the total number of lru entries in a list_lru_node in its nr_items field. In the case of memcg_drain_list_lru_node(), there is nothing to be done if nr_items is 0. We don't even need to take the nlru->lock as no new lru entry could be added by a racing list_lru_add() to the draining src_idx memcg at this point. Signed-off-by: Waiman Long <longman@redhat.com> --- mm/list_lru.c | 6 ++++++ 1 file changed, 6 insertions(+)