hw/nvme: reenable cqe batching
PATCH 1/1 hw/nvme: reenable cqe batching
From: Klaus Jensen <[email protected]>
Commit 2e53b0b45024 ("hw/nvme: Use ioeventfd to handle doorbell updates") had the unintended effect of disabling batching of CQEs.
This patch changes the sq/cq timers to bottom halfs and instead of calling nvme_post_cqes() immediately (causing an interrupt per cqe), we defer the call.
| iops
-----------------+------
baseline | 138k
+cqe batching | 233k
--- hw/nvme/ctrl.c | 26 +++++++++++--------------- hw/nvme/nvme.h | 4 ++-- 2 files changed, 13 insertions(+), 17 deletions(-)
hw/nvme/ctrl.c +11 -15
@@ -1401,13 +1401,7 @@ static void nvme_enqueue_req_completion(NvmeCQueue *cq, NvmeRequest *req)QTAILQ_REMOVE (&req->sq->out_req_list, req, entry);QTAILQ_INSERT_TAIL (&cq->req_list, req, entry); -if (req->sq->ioeventfd_enabled) { -/* Post CQE directly since we are in main loop thread */ -nvme_post_cqes (cq); - }else { -/* Schedule the timer to post CQE later since we are in vcpu thread */ -timer_mod (cq->timer,qemu_clock_get_ns (QEMU_CLOCK_VIRTUAL ) +500 ); - } +qemu_bh_schedule (cq->bh); }static void nvme_process_aers (void *opaque)
@@ -4252,7 +4246,7 @@ static void nvme_cq_notifier(EventNotifier *e)nvme_irq_deassert (n, cq); } -nvme_post_cqes (cq); +qemu_bh_schedule (cq->bh); }static int nvme_init_cq_ioeventfd (NvmeCQueue *cq)
@@ -4307,7 +4301,7 @@ static void nvme_free_sq(NvmeSQueue *sq, NvmeCtrl *n)uint16_t offset = sq->sqid <<3 ; n->sq[sq->sqid] =NULL ; -timer_free (sq->timer); +qemu_bh_delete (sq->bh);if (sq->ioeventfd_enabled) {memory_region_del_eventfd (&n->iomem,0x1000 + offset,4 , false,0 , &sq->notifier);
@@ -4381,7 +4375,8 @@ static void nvme_init_sq(NvmeSQueue *sq, NvmeCtrl *n, uint64_t dma_addr, sq->io_req[i].sq = sq;QTAILQ_INSERT_TAIL (&(sq->req_list), &sq->io_req[i], entry); } - sq->timer =timer_new_ns (QEMU_CLOCK_VIRTUAL , nvme_process_sq, sq); + + sq->bh =qemu_bh_new (nvme_process_sq, sq);if (n->dbbuf_enabled) { sq->db_addr = n->dbbuf_dbs + (sqid <<3 );
@@ -4698,7 +4693,7 @@ static void nvme_free_cq(NvmeCQueue *cq, NvmeCtrl *n)uint16_t offset = (cq->cqid <<3 ) + (1 <<2 ); n->cq[cq->cqid] =NULL ; -timer_free (cq->timer); +qemu_bh_delete (cq->bh);if (cq->ioeventfd_enabled) {memory_region_del_eventfd (&n->iomem,0x1000 + offset,4 , false,0 , &cq->notifier);
@@ -4771,7 +4766,7 @@ static void nvme_init_cq(NvmeCQueue *cq, NvmeCtrl *n, uint64_t dma_addr, } } n->cq[cqid] = cq; - cq->timer =timer_new_ns (QEMU_CLOCK_VIRTUAL , nvme_post_cqes, cq); + cq->bh =qemu_bh_new (nvme_post_cqes, cq); }static uint16_t nvme_create_cq (NvmeCtrl *n,NvmeRequest *req)
@@ -6913,9 +6908,9 @@ static void nvme_process_db(NvmeCtrl *n, hwaddr addr, int val)if (start_sqs) {NvmeSQueue *sq; QTAILQ_FOREACH(sq, &cq->sq_list, entry) { -timer_mod (sq->timer,qemu_clock_get_ns (QEMU_CLOCK_VIRTUAL ) +500 ); +qemu_bh_schedule (sq->bh); } -timer_mod (cq->timer,qemu_clock_get_ns (QEMU_CLOCK_VIRTUAL ) +500 ); +qemu_bh_schedule (cq->bh); }if (cq->tail == cq->head) {
@@ -6984,7 +6979,8 @@ static void nvme_process_db(NvmeCtrl *n, hwaddr addr, int val)pci_dma_write (&n->parent_obj, sq->db_addr, &sq->tail,sizeof (sq->tail)); } -timer_mod (sq->timer,qemu_clock_get_ns (QEMU_CLOCK_VIRTUAL ) +500 ); + +qemu_bh_schedule (sq->bh); } }
hw/nvme/nvme.h +2 -2
@@ -375,7 +375,7 @@ typedef struct NvmeSQueue {uint64_t dma_addr;uint64_t db_addr;uint64_t ei_addr; -QEMUTimer *timer; +QEMUBH *bh;EventNotifier notifier;bool ioeventfd_enabled;NvmeRequest *io_req;
@@ -396,7 +396,7 @@ typedef struct NvmeCQueue {uint64_t dma_addr;uint64_t db_addr;uint64_t ei_addr; -QEMUTimer *timer; +QEMUBH *bh;EventNotifier notifier;bool ioeventfd_enabled; QTAILQ_HEAD(, NvmeSQueue) sq_list;
Signature · 1 line
2.37.0 (Apple Git-136)
Quoted 8 lines · Klaus Jensen
From: Klaus Jensen <[email protected]>
Commit
2e53b0b45024("hw/nvme: Use ioeventfd to handle doorbell updates") had the unintended effect of disabling batching of CQEs.This patch changes the sq/cq timers to bottom halfs and instead of calling
nvme_post_cqes()immediately (causing an interrupt per cqe), we defer the call.
Nice change, looks good! Timers never did seem to be the best fit for this.
Quoted 6 lines · Klaus Jensen
Commit
2e53b0b45024("hw/nvme: Use ioeventfd to handle doorbell updates") had the unintended effect of disabling batching of CQEs.This patch changes the sq/cq timers to bottom halfs and instead of calling
nvme_post_cqes()immediately (causing an interrupt per cqe), we defer the call.
This change is definitely desired.
Quoted 5 lines
| iops
-----------------+------
baseline | 138k
+cqe batching | 233k
What is the queue depth config for this test?
Quoted 12 lines · Jinhao Fan
Quoted 6 lines · Klaus Jensen
Commit
2e53b0b45024("hw/nvme: Use ioeventfd to handle doorbell updates") had the unintended effect of disabling batching of CQEs.This patch changes the sq/cq timers to bottom halfs and instead of calling
nvme_post_cqes()immediately (causing an interrupt per cqe), we defer the call.This change is definitely desired.
Quoted 1 line
Quoted 7 lines
Quoted 4 lines
| iops
-----------------+------
baseline | 138k
+cqe batching | 233kWhat is the queue depth config for this test?
That's 64. My box isnt nearly as beefy as yours ;)
Jinhao Fan
Reviewed-byQuoted 12 lines · Klaus Jensen
Quoted 11 lines · Jinhao Fan
Quoted 6 lines · Klaus Jensen
Commit
2e53b0b45024("hw/nvme: Use ioeventfd to handle doorbell updates") had the unintended effect of disabling batching of CQEs.This patch changes the sq/cq timers to bottom halfs and instead of calling
nvme_post_cqes()immediately (causing an interrupt per cqe), we defer the call.This change is definitely desired.
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
| iops
-----------------+------
baseline | 138k
+cqe batching | 233kWhat is the queue depth config for this test?
That's 64. My box isnt nearly as beefy as yours ;)
:). The improvement looks great.
Quoted 9 lines · Klaus Jensen
From: Klaus Jensen <[email protected]>
Commit
2e53b0b45024("hw/nvme: Use ioeventfd to handle doorbell updates") had the unintended effect of disabling batching of CQEs.This patch changes the sq/cq timers to bottom halfs and instead of calling
nvme_post_cqes()immediately (causing an interrupt per cqe), we defer the call.
Quoted 7 lines
| iops
-----------------+------
baseline | 138k
+cqe batching | 233kFixes 2e53b0b45024 ("hw/nvme: Use ioeventfd to handle doorbell updates")Signed-off-by Klaus Jensen <[email protected]>
Thanks for the reviews, applied to nvme-next!