issues with 2.6 on desktop

phil-news-nospam_at_ipal.net
Date: 05/26/05 

Date: 26 May 2005 00:44:07 GMT

I switched to 2.6 on my desktop (have been running it OK on servers for a
while now) and a number of issues have started. Maybe you have some hints
where these can be resolved?

1. When doing strace of programs that spawn child processes and using the
    -f option of strace to trace forked processes, when a child exits, I
    get the message:
        PANIC: attached pid XXXXX exited
    and then everything is frozen solid. I cannot find the panic message
    in the kernel source, so I have no idea what is causing this. Could
    this be an strace incompatibility?

2. When I do a halt and shut down, the IDE drives are put into a state
    where they will not run when I reboot, and the reset button does not
    clear this state. Additionally, the power off button on the front
    powers off at least the video card and presumably the CPU, but the
    IDE drives and fans remain running. I have to power off using the
    AC disconnect switch in the back to clear everything. Then the IDE
    drives will be recognized by the BIOS (after a halt, even BIOS does
    see them as present) and the system can boot up.

    FYI, ACPI is supposedly supported by this hardware, but it has never
    worked even in 2.4 and even on other like machines that have been
    running 2.6 for a while. It could be an ACPI issue somewhere. Maybe
    the IDE drives are getting powered off?

3. At bootup, the first virtual console is OK, but all other virtual
    consoles have a funny cursor. The underline cursor is present with
    2 scan lines as usual, but the offset is such that the cursor is in
    middle of the character cell instead of the bottom. I can fix this
    with escape sequences and running SVGATextMode program. However,
    when X is run and I escape out of X back to virtual consoles, the
    problem is back again each time. However, if I set at least one of
    the virtual consoles to a block cursor, then escaping from X does NOT
    cause the problem to come back.

    It seems as though in 2.4 the cursor setting for the VC is used to set
    the register that controls the cursor in the video card each time any
    VC is changed. But it seems that 2.6 only does this if there is an
    actual change in VC which involves a change in per-VC setting. I am
    guessing that X has always goofed it up, but the 2.4 kernel corrected
    it due to the VC change while escaping from X. But 2.6 does not always
    set the cursor register, so the register gets left in the goofy state
    until some change would cause it to be set again.

4. Mouse speed in X is now twice as fast as before. Mouse speed in VC
    is not changed. This seems strange because I am using GPM's repeater
    mode and X is configured to use /dev/gpmdata for mouse input. When
    GPM is killed, the mouse won't move in X at all, so it seems to be
    configured as intended. But how the kernel could make the mouse move
    twice as fast in X as it moves in X under 2.4, both via GPM, while the
    VC speed is unchanged, is quite a mystery. 

-- 
-----------------------------------------------------------------------------
| Phil Howard KA9WGN       | http://linuxhomepage.com/      http://ham.org/ |
| (first name) at ipal.net | http://phil.ipal.org/   http://ka9wgn.ham.org/ |
-----------------------------------------------------------------------------
Quantcast