Re: Another 5.2-RC2 experience

Dan Strick wrote:
> I gave 5.2-RC2 a spin and found the following problems:
> 
> 1) Intel ICH5 SATA still doesn't work in native mode.  When I boot the
>    installation floppy diskettes each of my two SATA drisk drives draws
>    something in the range of 25-30 error messages of the general form:
> 
>         ad4: WARNING - READ DMA recovered from missing interrupt
> 
>    with a roughly 10 second pause before each message.  Someone suggested
>    that this was caused by a glitch in ACPI support, but this does not
>    seem to be the case.  I disabled the SATA controller and installed
>    5.2-RC2 on an old SCSI disk drive.  When I boot the installed 5.2-RC2
>    with the SATA controller reenabled, the missing-interrupt problem does
>    not always occur and is not suppressed by selecting the disabled-ACPI
>    boot option in the beastie bootstrap menu.  In fact, the SATA controller
>    may seem to work just fine with the normal boot and then fail when I
>    try the disabled-ACPI boot.  The ICH5 SATA controller seems to work
>    adequately well in both FreeBSD 4.9 and 5.1.

The ATA problems seem to be very hit-or-miss for different people with
identical hardware.  The best thing to do is contact Soeren Schmidt (the
ATA maintainer) and work with him to fix the problem.

> 
> 2) The gigabit ethernet controller on the i875P CSA bus (em driver),
>    consistently fails (apparently continuously timing out and resetting)
>    unless I boot with ACPI disabled (in which case it seems to work just
>    fine).  This device also seems to work correctly in both FreeBSD 4.9
>    and 5.1.
> 

5.2 is the first release ever of FreeBSD that attempts to correctly
route interrupts under ACPI.  For many people, it works flawlessly and
is mandatory for their hardware to work at all.  Unfortunatley, the ACPI
spec is often abused by the BIOS authors and our implementation cannot
work around all of the literally *hundreds* of different broken BIOSes.
The best thing to do here is to send John Baldwin the verbose dmesg
output both with and without ACPI, along with the output from the
mptable and acpidump commands.  In any case, I specifically wrote the
'beastie' bootloader so that disabling ACPI would be easy for these
kinds of situations.

> 3) The UFS1 file system super block summary data incompatibility (with
>    FreeBSD 4.9) is still a problem.

Please refresh my memory on this problem, I thought that all-known UFS
incompatibilities had been dealt with in some fashion.

Scott