5.2.1-RC2 debug kernel PANIC "Memory modified after free"

[Jesse Guardiani]

Jesse Guardiani wrote:

> Steve Kargl wrote:
> 
> [...]
> 
>>> > or setting it in
>>> > the loader.  The other workaround was proposed by Maxim.  You
>>> > need to change line 570 in dev/ata/ata-all.c from
>>> >    request->retries = -1;
>>> > to
>>> >    request->retries = 3;
>>> 
>>> Since I'm not using ACPI, are you sure about that?
>>> 
>> 
>> Works for me.  Although SOren committed a revision to ata-all.c
>> yesterday, which I haven't hade time to test.
>> 
>> http://lists.freebsd.org/pipermail/freebsd-current/2004-February/020580.html
>> http://lists.freebsd.org/pipermail/freebsd-current/2004-February/021323.html
>> http://lists.freebsd.org/pipermail/freebsd-current/2004-February/021500.html
> 
> Interesting. Well, I can confirm that the panic is definately NOT ACPI
> related. It happens in my debug kernel regardless of whether or not I have
> the acpi.ko module loaded and/or hint.acpi.0.disabled=1 set.
> 
> I don't have atapicam in the kernel in question either.
> 
> I'm going to try that patch from Maxim today. I'll report my findings.

Well, it took a bit longer than I expected to find the time to test Maxim's
patch properly.

Unfortunately, the short of it is that it did NOT solve my problem. Instead
of the kernel panicing after device detection, it now hangs DURING device
detection.

I have posted transcripts of my serial console sessions with normal AND verbose
boots AND ps AND trace outputs for the patched and non-patched kernels. Hopefully
an ATA guru like sos@ can take a look and give me some direction.

Here's the kernel config I used for each of these debug kernels:
    http://www.wingnet.net/~jesse/FreeBSD/TREVARTHAN-DEBUG

I completely deleted the kernel build directory before each kernel was config'd,
built, and installed.

And here's the patch that the boot.patched.txt kernels were built with:
    http://lists.freebsd.org/pipermail/freebsd-current/2004-February/021500.html


Transcripts:
=======================================

normal boot with 5.2.1rc2 debug kernel:
    http://www.wingnet.net/~jesse/FreeBSD/20040221-freebsd-5.2.1rc2-debug-normal-boot.txt

verbose boot with 5.2.1rc2 debug kernel:
    http://www.wingnet.net/~jesse/FreeBSD/20040221-freebsd-5.2.1rc2-debug-verbose-boot.txt


normal boot with 5.2.1rc2 debug kernel WITH Maxim's patch:
    http://www.wingnet.net/~jesse/FreeBSD/20040221-freebsd-5.2.1rc2-debug-normal-boot.patched.txt

verbose boot with 5.2.1rc2 debug kernel WITH Maxim's patch:
    http://www.wingnet.net/~jesse/FreeBSD/20040221-freebsd-5.2.1rc2-debug-verbose-boot.patched.txt


And to recap, the stock 5.2.1rc2 kernel boots just fine. However, it
core dumps on random ocassions, and I was hoping to get a gdb backtrace
from a debug kernel core. But my debug kernel won't boot!

The major differences I see from GENERIC to my debug kernel are that
I don't include 'options SMP' or 'options apic' in my debug kernel
and my debug kernel has WATCHDOG and all of the other debug options
turned on. I can post a diff if anyone is interested.

Thanks!

-- 
Jesse Guardiani, Systems Administrator
WingNET Internet Services,
P.O. Box 2605 // Cleveland, TN 37320-2605
423-559-LINK (v)  423-559-5145 (f)
http://www.wingnet.net