Re: init scripts problems ?

"Also sprach denis bonnenfant:"
> before echo 0 :
> 
> 5     0  2992     1  15   0  1992 1004 sync_p D    ?          0:00 enbd-client se3-croise 12355 -i A-2 -n 2 -m -b 512 /dev/nda

I wonder how one can get that WCHAN field wider :-). sync_pXXX ? At any
rate somebody is syncing something. It wuld be useful to know which
client daemon this is - master or slave. A pstree -p | grep enbd should
say.

I suspect it is the master.

> 1     0  3118     1  15   0     0    0 md_thr S    ?          0:00 [md0_raid1]
> 
> after echo 0 :
> 
> it works, this time...
> 
> md: error, md_import_device() returned -22
> ENBD #81[0]: enbd_notify_md_device ioctl 2344 to raid devices returned -22
> ENBD #2664[3]: enbd_clr_sock enabled device nda
> ENBD #2251[8]: enbd_clr_queue unqueued 1 reqs
> ENBD #2251[9]: enbd_clr_queue unqueued 0 reqs
> ENBD #3288[7]: enbd_soft_reset INVALIDATE DEVICE nda BEGIN
> ENBD #3292[0]: enbd_soft_reset invalidate would crash on device 00000000 with inode 00000000
> ENBD #3293[0]: enbd_soft_reset aborting invalidate

Well well - there is still serious breakage, but one is saved by the
test. lo->inode or lo->inode->bdev is NULL. lo->inode is NEVER unset by
enbd after setting it, so it must be lo->inode->bdev that is null. It
has probably been killed by a blkdev_get on the last opener.

I'll provide that openers patch in a moment.

Peter