I have added the 'Signed-off-by:' line.

after NVME_PROBE_IDENTIFY_NS is done, fire a AC_GETDEV_CHANGED async event.

Thank you!

update as reviewer suggested.

fix unrelated change.

rebase as reviewer suggested.

Any chance to get some notice? everytime I install a new machine, I have to modify fstab manually to avoid this issue.

Yes, please commit this for me. thank you.

update comment as suggested.

In D32657#749683, @wanpengqian_gmail.com wrote:

explicitly initialize MXPS and NOPS to zero as well

explicitly initialize MXPS and NOPS to zero as well

Add a B suffix to match the default one.

fix typo and modify words as comment suggested.

use humanize_number() instead of new wheel.

I don't think we need the namespace info before interrupts are enabled, so I don't think we need a kthread for this. At most, we may need a workqueue to run job...

Yes. Normally we only query namespaces while we're starting up (hence the poll complete). We
should let this be event driven instead.

panic: sleepq_add: td 0xfffffe00e228b020 to sleep on wchan 0xffffffff81684b85 with sleeping pro
cpuid = 5
time = 1637052864
KDB: stack backtrace:
db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 0xfffffe00e21e7b00
vpanic() at vpanic+0x181/frame 0xfffffe00e21e7b50
panic() at panic+0x43/frame 0xfffffe00e21e7bb0
sleepq_add() at sleepq_add+0x359/frame 0xfffffe00e21e7c00
_sleep() at _sleep+0x1f5/frame 0xfffffe00e21e7ca0
pause_sbt() at pause_sbt+0xfe/frame 0xfffffe00e21e7cd0
nvme_ns_construct() at nvme_ns_construct+0xd3/frame 0xfffffe00e21e7d70
nvme_notify_ns() at nvme_notify_ns+0x50/frame 0xfffffe00e21e7da0
nvme_ctrlr_async_event_log_page_cb() at nvme_ctrlr_async_event_log_page_cb+0xf6/frame 0xfffffe0
nvme_qpair_complete_tracker() at nvme_qpair_complete_tracker+0x20f/frame 0xfffffe00e21e7e10
nvme_qpair_process_completions() at nvme_qpair_process_completions+0x19f/frame 0xfffffe00e21e7e
ithread_loop() at ithread_loop+0x279/frame 0xfffffe00e21e7ef0
fork_exit() at fork_exit+0x80/frame 0xfffffe00e21e7f30
fork_trampoline() at fork_trampoline+0xe/frame 0xfffffe00e21e7f30
--- trap 0, rip = 0, rsp = 0, rbp = 0 ---
KDB: enter: panic
[ thread pid 12 tid 100098 ]
Stopped at      kdb_enter+0x37: movq    $0,0x1276e2e(%rip)

remove nvd modification.

filter out all the non NMVE_LOG_CHANGED_NAMESPACE

In D32953#745698, @imp wrote:

So if I take these patches, how do I change the namespace size in bhyve? I can fix nvd and nda

Yea, they need to be notified when the size changes, though, and only the nvme driver knows that.
There's a nvme_sim_ns_change callback that's currently registered for nda. This is the same as nvd_new_disk,
but I'm not sure that nvd_new_disk handles the change case well.
I'll have to thread this through. My current ns enhancement work is incomplete and doesn't handle this deatail.

https://reviews.freebsd.org/D32963 is the current state of things, but I don't see any calls to disk_resize() in it.
scsi and ata disks work,but nvme doesn't. I've not worked through all the details of that review, btw, since it
is a work in progress that had other things take priority over and I've not returned to it

add/correct comments and output DPRINTF log more detail.

Use more properly function name, update sytle.

In D32953#744004, @imp wrote:

In D32953#744003, @wanpengqian_gmail.com wrote:

The nvmecontrol of FreeBSD guest can output the changed size. but gpart still reports the original size. how can we let kernel reload this change? nvme reset seems no a fix.

rewrite the partition table?

what disk diskinfo ndaX or nvdX say?

The nvmecontrol of FreeBSD guest can output the changed size. but gpart still reports the original size. how can we let kernel reload this change? nvme reset seems no a fix.

kill goto.

fix typo.

correct the timestamp output.

use a more suitable name for macro.

fix the timestamp caculation.

update manpage as comments suggested.
also add print Timestamp for details.

create struct for each feature object

create struct for each feature object

create struct for each feature object

use function-like macro.

use function-like macro.

update manpage.

Add command examples.

fix typo.

update manpage

copy code from sys/dev/nvme/nvme_qpair.c,
move _SHIFT _MASK difinition to sys/dev/nvme/nvme.h, seperate ticket: D32885
remove these difinitions when D32885 is accepted.

use #define instead of numbers.

define more define and other updates.

update others too. (An nsid)

Fix grammer/comments as reviewer suggested.

create a macro to calculate power_on_hours. it can be shared with others.

In D32700#741111, @imp wrote:

One other question: Are there vendor specific features that might benefit from the extensible module loading we do for commands and mode pages?

Fix grammer/comments as reviewer suggested.

fix a typo.

rename function name to match other files's code style.

Print detail error message when failure.

add more print functions.

Add more print functions.

fix condition value.

Implement Set Feature for Temperature Threshold.