/dev permissions after reboot (and panic)
Peter Avalos
pavalos at theshell.com
Sat Jul 16 20:22:27 PDT 2005
On Sat, Jul 16, 2005 at 08:05:24PM -0700, Matthew Dillon wrote:
>
> :Content-Type: text/plain; charset=us-ascii
> :Content-Disposition: inline
> :
> :So every time I reboot, I get a panic (I believe someone else may
> :be having this problem too). I've included a capture of the ddb
> :session at the bottom. Unfortunately, I can't get a core dump.
> :But, what really sucks, is if I did a 'MAKEDEV all', permissions
> :in /dev get hosed when the machine comes back up. MAKEDEV is doing
> :the right thing, because I've checked to make sure the devices
> :have the correct permissions. When the machine comes back up,
> :things like null, zero, tty, lose their write permissions, which
> :hoses things up.
> :
> :A copy of the kernel can be found on leaf:~pavalos/public/kernel.0713
> :or at:
> :
> :http://www.theshell.com/~pavalos/crash/kernel.0713
> :
> :Here's the ddb session (hand-typed):
>
> Are you rebooting just after doing the MAKEDEV or are the permissions
> hosed after you run MAKEDEV and check them, before rebooting?
>
> If the permissions are getting hosed after the reboot see if a few
> 'sync' commands and a pause before 'reboot' works around that problem.
> I will try to reproduce the MAKEDEV issue on my test box.
>
Yeah, I've tried a bunch of sync's before the reboot. I've even seen this
happen hours after a MAKEDEV.
> Which patch set is this with? The m_nextpkt != NULL is a warning
> rather a panic, but the patch set I sent today should have fixed that.
> The actual panic is occuring due to the buffers not being flushable,
> which sounds like a softupdates issue so if the sync commands don't
> fix the problem (or even if they do). See if turning off softupdates
> fixes the problem (which will tell me where to look).
>
That was with no patches...just straight-up HEAD. I'm compiling up a
new kernel with the SOCKBUF_DEBUG patchset right now, and I'll see if
that warning goes away. I'll also try turning off softupdates and see
what happens.
--Peter
Attachment:
pgp00007.pgp
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pgp00007.pgp
Type: application/octet-stream
Size: 189 bytes
Desc: "Description: PGP signature"
URL: <http://lists.dragonflybsd.org/pipermail/bugs/attachments/20050716/c06222c7/attachment-0022.obj>
More information about the Bugs
mailing list