(2007-08-04)
SSL connect attempt failed with unknown errorerror:1408F10B:SSL routines:SSL3_GET_RECORD:wrong version number

	This can probably happen with any program using OpenSSL, and it most likely
	means that the peer didn't send a useful handshake at all. (like trying to connect
	an ssl client to a unencrypted server.) Ran into this when trying to use the Net::FTPSSL
	perl modules with the MOVEit ftps server. There's a bug in Net::FTPSSL that caused it to
	not finish reading the plaintext data on the socket before switching to SSL, which 
	apparently causes openssl's handshake code to choke on remaining plaintext data
	left in the receive buffer.

					* * *

(2007-08-15)
Tyan S2696 cpu & fbdimm temp and fan control with linux:

	Download ADT7470 patch from http://www.lm-sensors.org/wiki/Devices
	Optionally build the driver for W83627EHF as well. (mobo has 2 mon chips)
Show cpu temps:
	grep . /sys/bus/i2c/devices/0-002e/temp{2,3,5,6}_input
Show cpu fan speeds:
	grep . /sys/bus/i2c/devices/0-002e/fan1_input /sys/bus/i2c/devices/9191-0c00/fan2_input 
Set auto control from /etc/rc.d/rc.local:
	echo 0 >/sys/bus/i2c/devices/0-002e/pwm1_auto_point1_pwm	; safe. fans never stop.
	echo 50000 >/sys/bus/i2c/devices/0-002e/pwm1_auto_point1_temp	; or 50-55 degC or so
	echo 255 >/sys/bus/i2c/devices/0-002e/pwm1_auto_point2_pwm
	#echo 70000 >/sys/bus/i2c/devices/0-002e/pwm1_auto_point2_temp	; automatic
	echo 2 >/sys/bus/i2c/devices/0-002e/pwm1_auto_channels_temp	; $[1<<($sensor_number-1)]
	echo 2 >/sys/bus/i2c/devices/0-002e/pwm1_enable

	it looks like both cpu fans (dual cpu board) are controled by the
	same pwm channel(1). on a dual xeon 5160 (2x2 cores) it looks like this:
	/sys/bus/i2c/devices/0-002e/temp2_input: responds to all cores
	/sys/bus/i2c/devices/0-002e/temp3_input: responds to all cores
	/sys/bus/i2c/devices/0-002e/temp5_input: responds to core 0 and 2
	/sys/bus/i2c/devices/0-002e/temp6_input: responds to core 1 and 3
	(core numbers according to linux sched_setaffinity(), top, etc)

This appears to be needed to make the auto control actually do anything:
	/etc/inittab: fans:345:respawn:/bin/sh -c "while sleep 40; do grep x /sys/bus/i2c/devices/0-002e/temp2_input; done"

	The following program prints the current temperatore of the on-board sensors on
	each of the 8 FBDIMM sticks: fbdimm_temp.c (should stay below 85...)
	FB-DIMM 1 and 3 are the hottest because they don't get blown on by cpu fans.

	This motherboard also needs a bigger heatsink and/or fan on the north bridge.
	Try this one NC-U6

					* * *

(2007-09-19)
"ORA-12547: TNS:lost contact" when trying to connect to oracle after enabling
VLM buffer cache. (use_indirect_data_buffers=true)... if this error occurs
when trying to connect to oracle as users other than the oracle owner, it's
probably because /etc/security/limits.conf needs to allow all users connecting
to oracle to pin large amounts of memory (the vlm window). so try:

	* soft memlock 3145728
	* hard memlock 3145728

instead of

	oracle soft memlock 3145728
	oracle hard memlock 3145728

confirm with ulimit -l, or run ulimit -l 3145728 in a shell and reconnect.