Current code overestimates the value of max_outstanding_req_per_channel for Win8 and newer hosts, since vmscsi_size_delta is set to the initial value of sizeof(vmscsi_win8_extension) rather than zero. This may lead to wrong decisions when using ring_avail_percent_lowater equals to zero. The estimate of max_outstanding_req_per_channel is 'exact' for Win7 and older hosts. A better choice, keeping the algorithm for the estimation simple, is to err the other way around, i.e., to underestimate for Win7 and older but to use the exact value for Win8 and newer. Suggested-by: Dexuan Cui <decui@xxxxxxxxxxxxx> Signed-off-by: Andrea Parri (Microsoft) <parri.andrea@xxxxxxxxx> Cc: "James E.J. Bottomley" <jejb@xxxxxxxxxxxxx> Cc: "Martin K. Petersen" <martin.petersen@xxxxxxxxxx> Cc: linux-scsi@xxxxxxxxxxxxxxx --- drivers/scsi/storvsc_drv.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/storvsc_drv.c b/drivers/scsi/storvsc_drv.c index ded00a89bfc4e..64298aa2f151e 100644 --- a/drivers/scsi/storvsc_drv.c +++ b/drivers/scsi/storvsc_drv.c @@ -2141,12 +2141,15 @@ static int __init storvsc_drv_init(void) * than the ring buffer size since that page is reserved for * the ring buffer indices) by the max request size (which is * vmbus_channel_packet_multipage_buffer + struct vstor_packet + u64) + * + * The computation underestimates max_outstanding_req_per_channel + * for Win7 and older hosts because it does not take into account + * the vmscsi_size_delta correction to the max request size. */ max_outstanding_req_per_channel = ((storvsc_ringbuffer_size - PAGE_SIZE) / ALIGN(MAX_MULTIPAGE_BUFFER_PACKET + - sizeof(struct vstor_packet) + sizeof(u64) - - vmscsi_size_delta, + sizeof(struct vstor_packet) + sizeof(u64), sizeof(u64))); #if IS_ENABLED(CONFIG_SCSI_FC_ATTRS) -- 2.25.1
