Page MenuHomeFreeBSD
Differential D36345 Diff 110002 usr.sbin/bhyve/pci_nvme.c

Changeset View

Standalone View

View Options

usr.sbin/bhyve/pci_nvme.c

Show First 20 Lines • Show All 536 Lines • ▼ Show 20 Linespci_nvme_init_ctrldata(struct pci_nvme_softc *sc)
cd->mdts = NVME_MDTS; /* max data transfer size (2^mdts * CAP.MPSMIN) */ cd->mdts = NVME_MDTS; /* max data transfer size (2^mdts * CAP.MPSMIN) */
cd->ver = NVME_REV(1,4); cd->ver = NVME_REV(1,4);
cd->cntrltype = NVME_CNTRLTYPE_IO; cd->cntrltype = NVME_CNTRLTYPE_IO;
cd->oacs = 1 << NVME_CTRLR_DATA_OACS_FORMAT_SHIFT; cd->oacs = 1 << NVME_CTRLR_DATA_OACS_FORMAT_SHIFT;
cd->oaes = NVMEB(NVME_CTRLR_DATA_OAES_NS_ATTR); cd->oaes = NVMEB(NVME_CTRLR_DATA_OAES_NS_ATTR);
cd->acl = 2; cd->acl = 3;
cd->aerl = 4; cd->aerl = 3;
chuckUnsubmitted
Not Done
Inline Actions

The wording in the specification is:

It is recommended that implementations support a minimum of four Asynchronous
Event Request Limit commands outstanding simultaneously.

Supporting 5 (aerl = 4) would meet their recommendation, wouldn't you agree?

chuck: The wording in the specification is: > It is recommended that implementations support a…
wanpengqian_gmail.comAuthorUnsubmitted
Done
Inline Actions

Yes, a value higher than 4 is totally OK as specification says. but the value of 5 is a little un-normal. I don't know why we pick up this value at first. as 4 is the value that other controllers do as far as my observation.

wanpengqian_gmail.com: Yes, a value higher than 4 is totally OK as specification says. but the value of 5 is a little…
chuckUnsubmitted
Not Done
Inline Actions

There isn't a limit on the number of concurrent aborts in the implementation. So one arbitrary value is as good as another, but I'd suggest removing the comment.

chuck: There isn't a limit on the number of concurrent aborts in the implementation. So one arbitrary…
wanpengqian_gmail.comAuthorUnsubmitted
Done
Inline Actions

You are right, currently the Abort Command is not processed by the controller. I'd like the value to be more normal as real nvme controllers use the specification recommended value. I will remove the comment later.

wanpengqian_gmail.com: You are right, currently the Abort Command is not processed by the controller. I'd like the…
/* Advertise 1, Read-only firmware slot */ /* Advertise 1, Read-only firmware slot */
cd->frmw = NVMEB(NVME_CTRLR_DATA_FRMW_SLOT1_RO) | cd->frmw = NVMEB(NVME_CTRLR_DATA_FRMW_SLOT1_RO) |
(1 << NVME_CTRLR_DATA_FRMW_NUM_SLOTS_SHIFT); (1 << NVME_CTRLR_DATA_FRMW_NUM_SLOTS_SHIFT);
cd->lpa = 0; /* TODO: support some simple things like SMART */ cd->lpa = 0; /* TODO: support some simple things like SMART */
cd->elpe = 0; /* max error log page entries */ cd->elpe = 0; /* max error log page entries */
/* /*
* Report a single power state (zero-based value) * Report a single power state (zero-based value)
▲ Show 20 Lines • Show All 2,821 Lines • Show Last 20 Lines