biowait for seconds

[Emmanuel Dreyfus <manu%netbsd.org@localhost>]

Hello

This is the same machine that crashed before reaching single user, as
detailed here: https://mail-index.netbsd.org/tech-kern/2025/04/11/msg030349.html

With a helpful patch from riastradh@ it is now able to boot, but it exhibit
a strange behavior. Processes sometimes sleep for seconds in biowait.

Here is an example, tar stuck in wapbl_begin
trace: pid 17780 lid 17780 at 0xffffb2093b984c00
sleepq_block() at sleepq_block+0xee
turnstile_block() at turnstile_block+0x25c
rw_vector_enter() at rw_vector_enter+0x13e
wapbl_begin() at wapbl_begin+0x6c
ffs_write() at ffs_write+0x149
VOP_WRITE() at VOP_WRITE+0xeb
vn_write() at vn_write+0xf3
dofilewrite() at dofilewrite+0x80
sys_write() at sys_write+0x49
syscall() at syscall+0x9d

And ioflush in biowait. 
trace: pid 0 lid 1336 at 0xffffb20930dabcc0
sleepq_block() at sleepq_block+0xee
cv_wait() at cv_wait+0xca
biowait() at biowait+0x4f
wapbl_buffered_flush() at wapbl_buffered_flush+0xd8
wapbl_buffered_write() at wapbl_buffered_write+0x115
wapbl_circ_write() at wapbl_circ_write+0x101
wapbl_flush() at wapbl_flush+0x3cb
ffs_sync() at ffs_sync+0x1cd
VFS_SYNC() at VFS_SYNC+0x22
sched_sync() at sched_sync+0xdd

Mounting without -o log, things get even more strange. tar will hang 
at random time, and hitting a key on another terminal let it carry on.

This is NetBSD-current on amd64, underlying storage is a nvme board.

nvme0 at pci1 dev 0 function 0: Marvell product 2241 (rev. 0x20)
nvme0: NVMe 1.3
nvme0: interrupting at ioapic0 pin 16
nvme0: Dell BOSS-N1, firmware 11131069, serial VN0WW56VFCV004A801TY
ld0 at nvme0 nsid 1


-- 
Emmanuel Dreyfus
manu%netbsd.org@localhost