I've had a series of system hangs where I hoped to get a crash dump to at least get some clues but the normal "magic sysrq" functionality is broken too; not even the immediate reboot key works. The keyboard state LEDs (Caps Lock, etc) no longer toggle. The only thing that works is holding down the power button on the PC.
Some more background below, but my generic question is what the next steps for debugging such a hang would be, i.e. how can I get at least a callstack (presuming the culprit is still running) and preferably a working (mini) debugger against such an unresponsive state?
Details: it's a desktop PC (Dell OptiPlex 7010) running a fairly recent OS [stock LinuxMint 16, kernel 3.11.0-12-generic], with a PS/2 keyboard. It has a serial port, but unfortunately I don't have another machine + null modem cable handy to try to use for a kernel console. I (naively) suspect that if Alt-SysRq isn't working then attempts to switch desktops, use netconsole, etc, will be futile.
Also as supplementary info (probably not relevant), the hangs were from using fscache against an overloaded CIFS network mount. Caching was working (/proc/fs/fscache/stats showed hits happening, so the config wasn't completely wrong) but with this periodic "brutal hang". The kernel log does have the suggestive entry:
FS-Cache: Cookie type CIFS.uniqueid marked page 0 multiple times
for at least some of the hangs. Memory tests ok and the correlation between the hangs and using fscache is very solid, so I doubt it's a generic hardware issue (RAM, temperature, cosmic rays...)