2.4.16-9: answered "no" to everything except CONFIG_NET which you have to
have because something in ECN handling requires af_inet (in struct sock?)
So far it hasn't died.
    I did `while true; do sum /dev/mem; done` 24 times taking 427 secs (a
little over 7 mins) and it didn't freeze.  Plus setup and checkout beforehand.
This is the longest time kernel 2.4.16 has ever run.

2.4.16-10: Turning on (General Setup) hotplug; PCMCIA/Cardbus; SysV IPC.
(Networking options) UNIX domain sockets.
    It didn't freeze; I did `sum /dev/mem` as above.

2.4.16-11: Adding (General Setup) Power management; APM BIOS; CPU idle calls;
    console blanking.  Does not freeze in the above test.

2.4.16-12: Adding everything from 2.4.16-1 that was compiled as a module,
    except excluding sound, USB, video (AGP, DRM), joysticks, wireless,
    netfilter.  Does not freeze in the above test.

2.4.16-13: Add IP multicast, syncookies, IDE DMA.  Does not freeze.

2.4.16-14: Add "enable APM at boot time", framebuffer, char devices: watchdog,
    i8x0 random, RTC, /dev/agpart with Radeon, DRM.  Does not freeze.

2.4.16-15: Add core input support, sound, joysticks, netfilter.  
    Does not freeze.  

2.4.16-16: Add USB support.  Does not freeze.  

2.4.16-17: Add wireless support.  Does not freeze.  

2.4.16-18: Add MSR, MTRR, APIC, more modules.  FREEZES.

2.4.16-19: Remove CONFIG_X86_UP_APIC only (Intel APIC on uniprocessor).  Works.

diff -c /boot/config-2.4.16-{18,19}
*** /boot/config-2.4.16-18      Mon Dec 17 22:58:32 2001
--- /boot/config-2.4.16-19      Tue Dec 18 13:01:46 2001
***************
*** 60,68 ****
  # CONFIG_MATH_EMULATION is not set
  CONFIG_MTRR=y
  # CONFIG_SMP is not set
! CONFIG_X86_UP_APIC=y
! # CONFIG_X86_UP_IOAPIC is not set
! CONFIG_X86_LOCAL_APIC=y
  
  #
  # General setup
--- 60,66 ----
  # CONFIG_MATH_EMULATION is not set
  CONFIG_MTRR=y
  # CONFIG_SMP is not set
! # CONFIG_X86_UP_APIC is not set
  
  #
  # General setup

CONFIG_X86_LOCAL_APIC is used in : (skipping CONFIG_X86_UP_IOAPIC and all SMP)
./init/main.c
A lot of header files
./arch/i386/kernel/traps.c
./arch/i386/kernel/setup.c
./arch/i386/kernel/irq.c
./arch/i386/kernel/time.c
./arch/i386/kernel/i8259.c
./arch/i386/kernel/apic.c
	Author is identified as Ingo Molnar <mingo@redhat.com>
./arch/i386/kernel/pci-irq.c
./arch/i386/kernel/nmi.c


What to send in as bug report: 
boot.msg when it fails
.config when failing, and the above diff.
Machine configuration.

/sbin/update -s 1 -f 1 -0 100 -4 50 -5 100
parms in fs/buffer.c (lies told by man page and by update -d about parm labels)
defaults: 30 sec data, 5 sec nondata
0:    40 Max fraction of LRU list to examine for dirty blocks (percent)
1:     0 Max nbr of dirty blocks to write (unused)
2:     0 Num of clean buffers to be loaded (unused)
3:     0 Dirty block threshold for activating (unseud)
4:   500 Delay btn kupdate flushes (jiffies)
5:  3000 Time for data buffers to age before flushing (jiffies)
6:    60 Percent requiring synchronous bdflush
7:     0 Time buffer cache load average constant (unused)
8:     0 LAV ratio (used to determine threshold for buffer fratricide) (unused)