Re: 5.3-BETA1 panic with getnewbuf: locked buf

"Doug White" <dwhite_at_gumbysoft.com> wrote in message 
news:<20040829170534.V69068_at_carver.gumbysoft.com>...
On Thu, 26 Aug 2004, Kachun Lee wrote:
 >  >The server (E7501 single XEON 4G RAM) that I upgraded from 4.10 to
 > 5.3-BETA few days >ago panic'ed twice on 'getnewbuf: locked buf'. I
 > included the backtrace and dmesg below. I >searched the lists and found
 > several similar instances mentioned. Anything I can help to >find the
 > problem. Thanks in advance for any help.
 >
 > I forget to mention that I already rebuilt its kernel with 1.75 kern_lock.c
 > (from current).
1.75 should fix this problem. If it didn't, make sure you're booting the
right kernel and run a full fsck.  There is another attempt in rev 1.76
you might try too.

Hi,

Thanks for the info! Maybe, with the time line of events, you can see if I 
am on the right track...

1. 5.3-BETA1 cvsup'ed on Tue (kern_lock.c v1.74). Upgraded a server from 
4.10 to 5.3-BETA1. Server panic'ed at end of dmesg.

2. Wed morning... got help from list to add ACPI. Server came up with 
5.3-BETA1.

3. Wed ~6pm... kernel panic'ed with kmem_map too small. Increased 
VM_KMEM_SIZE_MAX from 512M (which ran fine with 4.10) to 768M.

3. Server panic'ed on Thur ~6pm with getnewbuf. I located the related 
messages on the list and bought in kern_lock.c v1.75 from current. Rebooted 
server without full fsck.

4. Server crashed at ~5am Fri... bought server back up and send this head 
message for help.

5. Later in the morning, found kern_lock.c v1.76 and built 5.3-BETA1 with 
it. Installed into server at around 2pm Fri, but did do full fsck this time 
before going into multi-user mode.

6. Server has run through the weekend.

My remained concern in this issue was the server did panic'ed with 
kern_lock.c v1.75. According to your reply, that probably because I did not 
do a full fsck before bringing the system up. This is still a concern 
because, as far as I can tell, v1.74.2.1 in BETA2 is the same as v1.75.

Another problem... I tried enable PAE this morning, but the server hanged 
right after the loader. I found in the archive that someone had similar 
problem few months back, but could not find the resolution...

   http://lists.freebsd.org/pipermail/freebsd-current/2003-October/011452.html

I did not try removing ACPI, as the server would panic without it (see my 
earlier message). Any insight on this issue.

Regards
Kachun Lee

PS: I just upgraded the server to 5.3-BETA2 without PAE. I will try PAE 
with 5.3-BETA2 tomorrow.