>>> I'm running kernel 2.1.125 on a K6. When I run the Gnome "panel" with
>>> the battery-charge monitor applet, it kills my PPP throughput; ifconfig
>>> reports frequent framing errors. Other battery-charge monitors
>>> (e.g. wmapm) cause somewhat less trouble.
>> The gnome module regularly asks the apm driver for the battery status.
>> The apm module disables interrupts while it calls into the APM BIOS.
>> If your APM bios is slow interrupts are disabled for a long time and then
>> serial interrupts are delayed until the fifo of the serial UART overflows -
>> result is framing errors.
>How often is often? My work on Toshiba's would indicate that once every
>three or four seconds is more than adequate. Put another way Toshiba's
>battery monitoring utilities only inquire of the battery status every
>three or four seconds, and Toshiba know a thing or two about laptops.
Evidently the Gnome monitor checks a lot more often. It seems to me
that once a minute would be good enough. Anyway, different laptops have
different latencies for this operation; evidently recent Mitac systems
have really, really slow APM functions. I would rather not miss *any*
interrupts, and so have given up on running battery monitors on my