diff options
| author |   James Smart <jsmart2021@gmail.com> | 2021-09-10 16:31:55 -0700 |
|---|---|---|
| committer |   Martin K. Petersen <martin.petersen@oracle.com> | 2021-09-14 23:33:21 -0400 |
| commit | d5ac69b332d8859d1f8bd5d4dee31f3267f6b0d2 (patch) | |
| tree | 537594016ac66f391e6c8d73b0b29f08cae5015c /drivers/scsi/lpfc | |
| parent | scsi: lpfc: Fix EEH support for NVMe I/O (diff) | |
| download | linux-dev-d5ac69b332d8859d1f8bd5d4dee31f3267f6b0d2.tar.xz linux-dev-d5ac69b332d8859d1f8bd5d4dee31f3267f6b0d2.zip | |
scsi: lpfc: Adjust bytes received vales during cmf timer interval
The newly added congestion mgmt framework is seeing unexpected congestion
FPINs and signals. In analysis, time values given to the adapter are not
at hard time intervals. Thus the drift vs the transfer count seen is
affecting how the framework manages things.
Adjust counters to cover the drift.
Link: https://lore.kernel.org/r/20210910233159.115896-11-jsmart2021@gmail.com
Co-developed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Diffstat (limited to 'drivers/scsi/lpfc')
| -rw-r--r-- | drivers/scsi/lpfc/lpfc_init.c | 16 |
1 files changed, 14 insertions, 2 deletions
diff --git a/drivers/scsi/lpfc/lpfc_init.c b/drivers/scsi/lpfc/lpfc_init.c index 3fd730613de3..d3a2281cb48b 100644 --- a/drivers/scsi/lpfc/lpfc_init.c +++ b/drivers/scsi/lpfc/lpfc_init.c @@ -5876,7 +5876,7 @@ lpfc_cmf_timer(struct hrtimer *timer) uint32_t io_cnt; uint32_t head, tail; uint32_t busy, max_read; - uint64_t total, rcv, lat, mbpi; + uint64_t total, rcv, lat, mbpi, extra; int timer_interval = LPFC_CMF_INTERVAL; uint32_t ms; struct lpfc_cgn_stat *cgs; @@ -5943,7 +5943,19 @@ lpfc_cmf_timer(struct hrtimer *timer) phba->hba_flag & HBA_SETUP) { mbpi = phba->cmf_last_sync_bw; phba->cmf_last_sync_bw = 0; - lpfc_issue_cmf_sync_wqe(phba, LPFC_CMF_INTERVAL, total); + extra = 0; + + /* Calculate any extra bytes needed to account for the + * timer accuracy. If we are less than LPFC_CMF_INTERVAL + * add an extra 3% slop factor, equal to LPFC_CMF_INTERVAL + * add an extra 2%. The goal is to equalize total with a + * time > LPFC_CMF_INTERVAL or <= LPFC_CMF_INTERVAL + 1 + */ + if (ms == LPFC_CMF_INTERVAL) + extra = div_u64(total, 50); + else if (ms < LPFC_CMF_INTERVAL) + extra = div_u64(total, 33); + lpfc_issue_cmf_sync_wqe(phba, LPFC_CMF_INTERVAL, total + extra); } else { /* For Monitor mode or link down we want mbpi * to be the full link speed |