| Message ID | 20230303082115.449-1-KaiShen@linux.alibaba.com (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net-next,v3] net/smc: Use percpu ref for wr tx reference | expand |
On Fri, Mar 03, 2023 at 08:21:15AM +0000, Kai wrote: > The refcount wr_tx_refcnt may cause cache thrashing problems among > cores and we can use percpu ref to mitigate this issue here. We > gain some performance improvement with percpu ref here on our > customized smc-r verion. Applying cache alignment may also mitigate > this problem but it seem more reasonable to use percpu ref here. > We can also replace wr_reg_refcnt with one percpu reference like > wr_tx_refcnt. > > redis-benchmark on smc-r with atomic wr_tx_refcnt: > SET: 525817.62 requests per second, p50=0.087 msec > GET: 570841.44 requests per second, p50=0.087 msec > > redis-benchmark on the percpu_ref version: > SET: 539956.81 requests per second, p50=0.087 msec > GET: 587613.12 requests per second, p50=0.079 msec Does the test data need to be refreshed? > > Signed-off-by: Kai <KaiShen@linux.alibaba.com> > > v1->v2: > - Modify patch prefix > > v2->v3: > - Make wr_reg_refcnt a percpu one as well > - Init percpu ref with 0 flag instead of ALLOW_REINIT flag > --- > net/smc/smc_core.h | 10 ++++++++-- > net/smc/smc_wr.c | 35 ++++++++++++++++++++++++++++------- > net/smc/smc_wr.h | 5 ++--- > 3 files changed, 38 insertions(+), 12 deletions(-) > > diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h > index 08b457c2d294..1645fba0d2d3 100644 > --- a/net/smc/smc_core.h > +++ b/net/smc/smc_core.h > @@ -106,7 +106,10 @@ struct smc_link { > unsigned long *wr_tx_mask; /* bit mask of used indexes */ > u32 wr_tx_cnt; /* number of WR send buffers */ > wait_queue_head_t wr_tx_wait; /* wait for free WR send buf */ > - atomic_t wr_tx_refcnt; /* tx refs to link */ > + struct { > + struct percpu_ref wr_tx_refs; > + } ____cacheline_aligned_in_smp; > + struct completion tx_ref_comp; For the variable names suffixed with wr_*_refs, should we continue to use wr_*_refcnt? Thanks. > > struct smc_wr_buf *wr_rx_bufs; /* WR recv payload buffers */ > struct ib_recv_wr *wr_rx_ibs; /* WR recv meta data */ > @@ -122,7 +125,10 @@ struct smc_link { > > struct ib_reg_wr wr_reg; /* WR register memory region */ > wait_queue_head_t wr_reg_wait; /* wait for wr_reg result */ > - atomic_t wr_reg_refcnt; /* reg refs to link */ > + struct { > + struct percpu_ref wr_reg_refs; > + } ____cacheline_aligned_in_smp; > + struct completion reg_ref_comp; > enum smc_wr_reg_state wr_reg_state; /* state of wr_reg request */ > > u8 gid[SMC_GID_SIZE];/* gid matching used vlan id*/ > diff --git a/net/smc/smc_wr.c b/net/smc/smc_wr.c > index b0678a417e09..0021065a600a 100644 > --- a/net/smc/smc_wr.c > +++ b/net/smc/smc_wr.c > @@ -377,12 +377,11 @@ int smc_wr_reg_send(struct smc_link *link, struct ib_mr *mr) > if (rc) > return rc; > > - atomic_inc(&link->wr_reg_refcnt); > + percpu_ref_get(&link->wr_reg_refs); > rc = wait_event_interruptible_timeout(link->wr_reg_wait, > (link->wr_reg_state != POSTED), > SMC_WR_REG_MR_WAIT_TIME); > - if (atomic_dec_and_test(&link->wr_reg_refcnt)) > - wake_up_all(&link->wr_reg_wait); > + percpu_ref_put(&link->wr_reg_refs); > if (!rc) { > /* timeout - terminate link */ > smcr_link_down_cond_sched(link); > @@ -647,8 +646,10 @@ void smc_wr_free_link(struct smc_link *lnk) > smc_wr_wakeup_tx_wait(lnk); > > smc_wr_tx_wait_no_pending_sends(lnk); > - wait_event(lnk->wr_reg_wait, (!atomic_read(&lnk->wr_reg_refcnt))); > - wait_event(lnk->wr_tx_wait, (!atomic_read(&lnk->wr_tx_refcnt))); > + percpu_ref_kill(&lnk->wr_reg_refs); > + wait_for_completion(&lnk->reg_ref_comp); > + percpu_ref_kill(&lnk->wr_tx_refs); > + wait_for_completion(&lnk->tx_ref_comp); > > if (lnk->wr_rx_dma_addr) { > ib_dma_unmap_single(ibdev, lnk->wr_rx_dma_addr, > @@ -847,6 +848,20 @@ void smc_wr_add_dev(struct smc_ib_device *smcibdev) > tasklet_setup(&smcibdev->send_tasklet, smc_wr_tx_tasklet_fn); > } > > +static void smcr_wr_tx_refs_free(struct percpu_ref *ref) > +{ > + struct smc_link *lnk = container_of(ref, struct smc_link, wr_tx_refs); > + > + complete(&lnk->tx_ref_comp); > +} > + > +static void smcr_wr_reg_refs_free(struct percpu_ref *ref) > +{ > + struct smc_link *lnk = container_of(ref, struct smc_link, wr_reg_refs); > + > + complete(&lnk->reg_ref_comp); > +} > + > int smc_wr_create_link(struct smc_link *lnk) > { > struct ib_device *ibdev = lnk->smcibdev->ibdev; > @@ -890,9 +905,15 @@ int smc_wr_create_link(struct smc_link *lnk) > smc_wr_init_sge(lnk); > bitmap_zero(lnk->wr_tx_mask, SMC_WR_BUF_CNT); > init_waitqueue_head(&lnk->wr_tx_wait); > - atomic_set(&lnk->wr_tx_refcnt, 0); > + rc = percpu_ref_init(&lnk->wr_tx_refs, smcr_wr_tx_refs_free, 0, GFP_KERNEL); > + if (rc) > + goto dma_unmap; > + init_completion(&lnk->tx_ref_comp); > init_waitqueue_head(&lnk->wr_reg_wait); > - atomic_set(&lnk->wr_reg_refcnt, 0); > + rc = percpu_ref_init(&lnk->wr_reg_refs, smcr_wr_reg_refs_free, 0, GFP_KERNEL); > + if (rc) > + goto dma_unmap; > + init_completion(&lnk->reg_ref_comp); > init_waitqueue_head(&lnk->wr_rx_empty_wait); > return rc; > > diff --git a/net/smc/smc_wr.h b/net/smc/smc_wr.h > index 45e9b894d3f8..f3008dda222a 100644 > --- a/net/smc/smc_wr.h > +++ b/net/smc/smc_wr.h > @@ -63,14 +63,13 @@ static inline bool smc_wr_tx_link_hold(struct smc_link *link) > { > if (!smc_link_sendable(link)) > return false; > - atomic_inc(&link->wr_tx_refcnt); > + percpu_ref_get(&link->wr_tx_refs); > return true; > } > > static inline void smc_wr_tx_link_put(struct smc_link *link) > { > - if (atomic_dec_and_test(&link->wr_tx_refcnt)) > - wake_up_all(&link->wr_tx_wait); > + percpu_ref_put(&link->wr_tx_refs); > } > > static inline void smc_wr_drain_cq(struct smc_link *lnk) > -- > 2.31.1
On 3/8/23 4:30 PM, Tony Lu wrote: >> redis-benchmark on smc-r with atomic wr_tx_refcnt: >> SET: 525817.62 requests per second, p50=0.087 msec >> GET: 570841.44 requests per second, p50=0.087 msec >> >> redis-benchmark on the percpu_ref version: >> SET: 539956.81 requests per second, p50=0.087 msec >> GET: 587613.12 requests per second, p50=0.079 msec > > Does the test data need to be refreshed? > Will do. >> diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h >> index 08b457c2d294..1645fba0d2d3 100644 >> --- a/net/smc/smc_core.h >> +++ b/net/smc/smc_core.h >> @@ -106,7 +106,10 @@ struct smc_link { >> unsigned long *wr_tx_mask; /* bit mask of used indexes */ >> u32 wr_tx_cnt; /* number of WR send buffers */ >> wait_queue_head_t wr_tx_wait; /* wait for free WR send buf */ >> - atomic_t wr_tx_refcnt; /* tx refs to link */ >> + struct { >> + struct percpu_ref wr_tx_refs; >> + } ____cacheline_aligned_in_smp; >> + struct completion tx_ref_comp; > > For the variable names suffixed with wr_*_refs, should we continue to > use wr_*_refcnt? > > Thanks. In my opinion, we can't get the count of the percpu reference until it we start to destroy it. So maybe using wr_*_refcnt here is more appropriate？
On 3/13/23 9:20 AM, Kai wrote: > > > On 3/8/23 4:30 PM, Tony Lu wrote: >>> redis-benchmark on smc-r with atomic wr_tx_refcnt: >>> SET: 525817.62 requests per second, p50=0.087 msec >>> GET: 570841.44 requests per second, p50=0.087 msec >>> >>> redis-benchmark on the percpu_ref version: >>> SET: 539956.81 requests per second, p50=0.087 msec >>> GET: 587613.12 requests per second, p50=0.079 msec >> >> Does the test data need to be refreshed? >> > Will do. >>> diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h >>> index 08b457c2d294..1645fba0d2d3 100644 >>> --- a/net/smc/smc_core.h >>> +++ b/net/smc/smc_core.h >>> @@ -106,7 +106,10 @@ struct smc_link { >>> unsigned long *wr_tx_mask; /* bit mask of used >>> indexes */ >>> u32 wr_tx_cnt; /* number of WR send buffers */ >>> wait_queue_head_t wr_tx_wait; /* wait for free WR send >>> buf */ >>> - atomic_t wr_tx_refcnt; /* tx refs to link */ >>> + struct { >>> + struct percpu_ref wr_tx_refs; >>> + } ____cacheline_aligned_in_smp; >>> + struct completion tx_ref_comp; >> >> For the variable names suffixed with wr_*_refs, should we continue to >> use wr_*_refcnt? >> >> Thanks. > In my opinion, we can't get the count of the percpu reference until it > we start to destroy it. So maybe using wr_*_refcnt here is more > appropriate？ I mean wr_*_refs here. Sorry for the mistake.
diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index 08b457c2d294..1645fba0d2d3 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -106,7 +106,10 @@ struct smc_link { unsigned long *wr_tx_mask; /* bit mask of used indexes */ u32 wr_tx_cnt; /* number of WR send buffers */ wait_queue_head_t wr_tx_wait; /* wait for free WR send buf */ - atomic_t wr_tx_refcnt; /* tx refs to link */ + struct { + struct percpu_ref wr_tx_refs; + } ____cacheline_aligned_in_smp; + struct completion tx_ref_comp; struct smc_wr_buf *wr_rx_bufs; /* WR recv payload buffers */ struct ib_recv_wr *wr_rx_ibs; /* WR recv meta data */ @@ -122,7 +125,10 @@ struct smc_link { struct ib_reg_wr wr_reg; /* WR register memory region */ wait_queue_head_t wr_reg_wait; /* wait for wr_reg result */ - atomic_t wr_reg_refcnt; /* reg refs to link */ + struct { + struct percpu_ref wr_reg_refs; + } ____cacheline_aligned_in_smp; + struct completion reg_ref_comp; enum smc_wr_reg_state wr_reg_state; /* state of wr_reg request */ u8 gid[SMC_GID_SIZE];/* gid matching used vlan id*/ diff --git a/net/smc/smc_wr.c b/net/smc/smc_wr.c index b0678a417e09..0021065a600a 100644 --- a/net/smc/smc_wr.c +++ b/net/smc/smc_wr.c @@ -377,12 +377,11 @@ int smc_wr_reg_send(struct smc_link *link, struct ib_mr *mr) if (rc) return rc; - atomic_inc(&link->wr_reg_refcnt); + percpu_ref_get(&link->wr_reg_refs); rc = wait_event_interruptible_timeout(link->wr_reg_wait, (link->wr_reg_state != POSTED), SMC_WR_REG_MR_WAIT_TIME); - if (atomic_dec_and_test(&link->wr_reg_refcnt)) - wake_up_all(&link->wr_reg_wait); + percpu_ref_put(&link->wr_reg_refs); if (!rc) { /* timeout - terminate link */ smcr_link_down_cond_sched(link); @@ -647,8 +646,10 @@ void smc_wr_free_link(struct smc_link *lnk) smc_wr_wakeup_tx_wait(lnk); smc_wr_tx_wait_no_pending_sends(lnk); - wait_event(lnk->wr_reg_wait, (!atomic_read(&lnk->wr_reg_refcnt))); - wait_event(lnk->wr_tx_wait, (!atomic_read(&lnk->wr_tx_refcnt))); + percpu_ref_kill(&lnk->wr_reg_refs); + wait_for_completion(&lnk->reg_ref_comp); + percpu_ref_kill(&lnk->wr_tx_refs); + wait_for_completion(&lnk->tx_ref_comp); if (lnk->wr_rx_dma_addr) { ib_dma_unmap_single(ibdev, lnk->wr_rx_dma_addr, @@ -847,6 +848,20 @@ void smc_wr_add_dev(struct smc_ib_device *smcibdev) tasklet_setup(&smcibdev->send_tasklet, smc_wr_tx_tasklet_fn); } +static void smcr_wr_tx_refs_free(struct percpu_ref *ref) +{ + struct smc_link *lnk = container_of(ref, struct smc_link, wr_tx_refs); + + complete(&lnk->tx_ref_comp); +} + +static void smcr_wr_reg_refs_free(struct percpu_ref *ref) +{ + struct smc_link *lnk = container_of(ref, struct smc_link, wr_reg_refs); + + complete(&lnk->reg_ref_comp); +} + int smc_wr_create_link(struct smc_link *lnk) { struct ib_device *ibdev = lnk->smcibdev->ibdev; @@ -890,9 +905,15 @@ int smc_wr_create_link(struct smc_link *lnk) smc_wr_init_sge(lnk); bitmap_zero(lnk->wr_tx_mask, SMC_WR_BUF_CNT); init_waitqueue_head(&lnk->wr_tx_wait); - atomic_set(&lnk->wr_tx_refcnt, 0); + rc = percpu_ref_init(&lnk->wr_tx_refs, smcr_wr_tx_refs_free, 0, GFP_KERNEL); + if (rc) + goto dma_unmap; + init_completion(&lnk->tx_ref_comp); init_waitqueue_head(&lnk->wr_reg_wait); - atomic_set(&lnk->wr_reg_refcnt, 0); + rc = percpu_ref_init(&lnk->wr_reg_refs, smcr_wr_reg_refs_free, 0, GFP_KERNEL); + if (rc) + goto dma_unmap; + init_completion(&lnk->reg_ref_comp); init_waitqueue_head(&lnk->wr_rx_empty_wait); return rc; diff --git a/net/smc/smc_wr.h b/net/smc/smc_wr.h index 45e9b894d3f8..f3008dda222a 100644 --- a/net/smc/smc_wr.h +++ b/net/smc/smc_wr.h @@ -63,14 +63,13 @@ static inline bool smc_wr_tx_link_hold(struct smc_link *link) { if (!smc_link_sendable(link)) return false; - atomic_inc(&link->wr_tx_refcnt); + percpu_ref_get(&link->wr_tx_refs); return true; } static inline void smc_wr_tx_link_put(struct smc_link *link) { - if (atomic_dec_and_test(&link->wr_tx_refcnt)) - wake_up_all(&link->wr_tx_wait); + percpu_ref_put(&link->wr_tx_refs); } static inline void smc_wr_drain_cq(struct smc_link *lnk)
The refcount wr_tx_refcnt may cause cache thrashing problems among cores and we can use percpu ref to mitigate this issue here. We gain some performance improvement with percpu ref here on our customized smc-r verion. Applying cache alignment may also mitigate this problem but it seem more reasonable to use percpu ref here. We can also replace wr_reg_refcnt with one percpu reference like wr_tx_refcnt. redis-benchmark on smc-r with atomic wr_tx_refcnt: SET: 525817.62 requests per second, p50=0.087 msec GET: 570841.44 requests per second, p50=0.087 msec redis-benchmark on the percpu_ref version: SET: 539956.81 requests per second, p50=0.087 msec GET: 587613.12 requests per second, p50=0.079 msec Signed-off-by: Kai <KaiShen@linux.alibaba.com> v1->v2: - Modify patch prefix v2->v3: - Make wr_reg_refcnt a percpu one as well - Init percpu ref with 0 flag instead of ALLOW_REINIT flag --- net/smc/smc_core.h | 10 ++++++++-- net/smc/smc_wr.c | 35 ++++++++++++++++++++++++++++------- net/smc/smc_wr.h | 5 ++--- 3 files changed, 38 insertions(+), 12 deletions(-)