tmpfs panic: tmpfs_dir_getdotdotdent() at netbsd:tmpfs_dir_getdotdotdent+0x8b

[Thomas Klausner <wiz%NetBSD.org@localhost>]

Hi!

In my unvoluntary search for tmpfs bugs (see PRs 42484, 47739, 47480),
I found another one (I think) yesterday.

panic: trap
cpu9: Begin traceback...
vpanic() at netbsd:vpanic+0x1e3
panic() at netbsd:panic+0x43
trap() at netbsd:trap+0xd0c
--- trap (number 6) ---
tmpfs_dir_getdotdotdent() at netbsd:tmpfs_dir_getdotdotdent+0x8b
tmpfs_readdir() at netbsd:tmpfs_readdir+0x113
VOP_READDIR() at netbsd:VOP_READDIR+0x53
vn_readdir() at netbsd:vn_readdir+0x21b
sys___getdents30() at netbsd:sys___getdents30+0x60
syscall() at netbsd:syscall+0xb5
--- syscall (number 390) ---
7f7ff6d08f6a:
cpu9: End traceback...


I think this was triggered the following way: In my pkgsrc sandbox on
tmpfs, I had backgrounded a configure run (for graphics/graphviz).
Some time later, when the workdir (sandbox/scratch/graphics/graphviz)
and probably also it's parent dir (sandbox/scratch/graphics) were
removed, because I had finished working on graphviz, I foregrounded
the process again, the 'configure' continued and then the machine
paniced.

For perhaps understandable reasons I wasn't very motivated to retry
how reproducible it was.

Should I file a PR or is that probably one of the others in new form?

 Thomas