On Mon, 03 Oct 2005 16:01:08 -0700,
Steven Dake <[EMAIL PROTECTED]> wrote:
>Our QA found this problem with KDB:
>1. The target boots with KDB enabled kernel.
>2. Get KDB prompt by hitting ESC-KDB
>3. Set breakpoint (at sys_chmod)
>4. "go" on KDB prompt to come out.
>5. Verify that chmod invokes KDB.
>6. Disable KDB by echo 0 > /proc/sys/kernel/kdb
>7. Do chmod again. This results in segmentation fault.
>
>Please note that the system does not crash at this stage. However, if KDB is
>enabled again and invoked (by ESC-KDB), it does so with message "Catastrophic
>error detected".
>
>I've inlined a patch which fixes the problem by clearing any breakpoints when
>kdb is disabled.
The problem needs to be fixed, but the patch does it the wrong way.
Clearing all break points should be a simple 'kdb_parse("bc *\n", NULL);'.
Also the kdb hook into sysctl is no longer in kernel/sysctl.c, it has
been dynamic since 2.6.12-rc1.
---------------------------
Use http://oss.sgi.com/ecartis to modify your settings or to unsubscribe.
