Make kmod force default handlers before executing the user process.

into kernel.bkbits.net:/home/davem/net-2.5

the lane spec says the following about path switching:

10.1.2.7 Transmitting Held Frames

Once the LE_FLUSH_RESPONSE is received, the LE Client MUST
transmit any held  data frames on the new data path before
transmitting any further frames on the new path.

10.1.2.8 Switching Over Paths Without Flush

Regardless of the provisions of the rest of Section 10.1, when
switching from the old path to a new path, if an LE Client has
not transmitted a data frame to a particular LAN Destination via
the old path for a period of time greater than or equal to the
C22 Path Switching Delay, then it MAY start using the new path
without employing the Flush protocol.

the lane client in linux-atm does do this but it doesnt send the
held frames until a frame is transmitted on the vcc after the flush
has completed.  often this doesnt happen until you get an ip retransmit.
the following patch sends the held frames after the flush completes
(or the switching delay expires)

Stelian Pop reported that all signals are blocked in processes
execed from the kernel as usermode helpers.

 - suspend & remove for pci_socket was broken -- thanks to Paul
   Mackerras for noting this
 - to correctly initialize multiple pci_socket devices, a sock_offset is
   needed.
 - s_info doesn't need to be an array.

 - switch the Intel x86 speedstep/Pentium 4/Xeon drivers and the
   Transmeta Crusoe Longrun driver to use the advanced cpufreq_driver
   registration process
 - cleanups
 - spelling fixes (Petri Koistinen) - thanks!

Move x86 CPU_FREQ config choices to extra file & menu. (Marc-Christian Petersen)

 - Make MTRR driver compatible with 2.4/x86-64 again.  This fixes an
   endless loop in the XFree86 4.3pre server.
 - Fix the boot code that rejected earlier: rewrote early CPU detection
   and SSE is forced now
 - Always force inlining in vsyscalls

We should probably always read the vendor ID from each function
rather than assuming that it is identical to function 0.

cb_alloc() allocated a pci_dev on the stack to access PCI space.  This
is unnecessary since we have pci_bus_*_config_* functions.  Use these
functions instead.

cardbus.c uses its own names for our PCI config space functions.
Make it use our names.

read_cb_mem is only ever called with its "fn" argument set to zero.
We therefore do not need to pass it.

cb_setup_cis_mem doesn't reference the pci device.  Remove this unused
argument.

cb_config() and cb_release() are just simple dummy functions that are
only used in the internals of the PCMCIA code.  We inline them where
used.

into home.transmeta.com:/home/torvalds/v2.5/linux

Fixes a typo in pnp_check_*_conflicts functions.  Without this fix the
resource algorithm will work but will take longer to assign resources.

Also contains some minor reordering in pnp_activate_dev.

Reduces the stack memory usage in the following PnP Functions:
pnp_printf
pnp_set_current_resources
pnp_manual_config_dev
pnp_activate_dev

into kanga.twiddle.net:/home/rth/linux/axp-2.5

Hi Richard.

I noticed you have set -Werror in all alpha architecture
specific makefiles - good initiative. I wish we could do that in general.

But the correct way to alter CFLAGS in kbuild makefiles is to use
EXTRA_CFLAGS.
The following patch fixes this.
You could argue that this is a stylistic issue, but in this way
you follow Documentation/kbuild/makefiles.txt

I stumbled over:
EXTRA_AFLAGS := $(CFLAGS)
I could not realise the purpose with this assignment, and I noticed
that it did not include -Werror, becuse you did hat assignment in
the following line.
Wanted to inform you about it but I see no problem. It is just that
I did not understand it.

In math-emu I see this:
EXTRA_CFLAGS := -Iinclude/math-emu -w
Grepping the src I see no reason to have the -Iinclude/math-emu
Relevant files are prefixed with math-emu (soft-fp, single, double)
I did not fix it, because I cannot compile alpha.

	Sam (the Makefile nitpicker)

===== arch/alpha/kernel/Makefile 1.24 vs edited =====

x86-64 vsyscalls require mapping the sequence number used by
gettimeofday in a magic way, so that userland can access it via
vsyscalls for user space time-of-day access.

Instead of putting the magic into generic code I just allowed to move it
into architecture specific files.

Fix some annoying warnings that happen on x86-64.

 - Predeclare pt_regs in binfmts.h.  For some reason it isn't included
   implicitely there.
 - use unsigned long, not long for interrupt flags in posix-timers.c

This still has a few problems; e.g. posix-timers.c seems to be miscompiled
by many x86-64 compilers when compiled with -O1. But overall it is better
than before.

 - Disable numa again that was enabled by mistake
 - Fix CPU detection in early boot. Force SSE when needed.
 - Hack for nested NMIs.
 - Fix typo in vsyscall code that lead to broken time reporting
 - Merge with 2.5.62-bk4 (new switch_to, printk cleanups, new timer calls etc.)
 - Fix bug in strace - don't assume syscall_trace doesn't clobber %rcx
 - Fix gs load workaround
 - Fix return value of ioperm (found by new LTP test)
 - Remove some unecessary boot output
 - Remove some obsolete code in setup.c
 - Report more in /proc/cpuinfo, fix extended cpu family computation
 - Fix thread_info access from interrupt/exception stacks - this
   fixes many problems that especially showed on SMP (lost reschedule
   interrupts etc.)
 - Fix UP compilation
 - Symlink hugetlbpage.c from i386
 - Fix buggy lazy fpu handling

Older compilers require that automatic variable declarations appear at the
start of a block.