Paul Mackerras [Mon, 14 Nov 2005 06:31:55 +0000 (17:31 +1100)]
powerpc: Mark PREP and embedded as broken for now
These machines don't have working ARCH=powerpc support yet, so make
them depend on BROKEN so people don't enable them inadvertently and
get compile errors.
[PATCH] powerpc: kill ppc64 rtc.c, use genrtc instead
This moves the rtas RTC callbacks to rtas-rtc.c in arch/powerpc/kernel,
and kills the rest of arch/ppc64/kernel/rtc.c which was just a duplicate
of the genrtc functionality. Also enable build of genrtc for
CONFIG_PPC64 (it just works are we already have the required callbacks)
and enable it in all defconfigs.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
This fixes various errors in the new functions added in the vDSO's,
I've now verified all functions on both 32 and 64 bits vDSOs. It also
fix a sign extension bug getting the initial time of day at boot that
could cause the monotonic clock value to be completely on bogus for
64 bits applications (with either the vDSO or the syscall) on
powermacs.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
This patch uses a FORCE dependency on the arch/powerpc/include/asm
symlink so that it always gets rebuilt, thus avoiding all sort of funny
errors if the .config is changed between 32 and 64 bits.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
Michael Ellerman [Fri, 11 Nov 2005 13:06:05 +0000 (00:06 +1100)]
[PATCH] powerpc: Turn cpu_irq_down into kexec_cpu_down
We currently have a ppc_md member called cpu_irq_down, which disables IRQs
for the cpu in question. The only caller of cpu_irq_down is the kexec code.
On pSeries we need to do more than just teardown IRQs at kexec time, so rename
the ppc_md member to kexec_cpu_down and expand it. The pSeries code needs to
know, and other platforms might too, whether we're doing a crash shutdown (ie.
panicking) or a regular kexec, so add a flag for that.
The pSeries implementation of kexec_cpu_down does an unregister VPA call, which
tells the Hypervisor to stop writing stuff into our pacas. Without this we can
get weird memory corruption bugs when we kexec, caused by the Hypervisor
writing into the first kernel's pacas which happens to be somewhere interesting
in the second kernel's memory.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: Paul Mackerras <paulus@samba.org>
Michael Ellerman [Fri, 11 Nov 2005 03:25:24 +0000 (14:25 +1100)]
[PATCH] powerpc: Merge page.h
Merge asm-ppc/page.h and asm-ppc64/page.h into asm-powerpc/page.h,
asm-powerpc/page_32.h and asm-powerpc/page_64.h
Built for PPC (common_defconfig), with ARCH=powerpc, mostly built with
ARCH=ppc (other things break the build). Built and booted on P5 LPAR
for PPC64 with ARCH=ppc/powerpc (pseries_defconfig). Mostly built for
iSeries powerpc.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: Paul Mackerras <paulus@samba.org>
Martin Waitz [Mon, 14 Nov 2005 00:08:13 +0000 (16:08 -0800)]
[PATCH] DocBook: allow to mark structure members private
Many structures contain both an internal part and one which is part of the API
to other modules. With this patch it is possible to only include these public
members in the kernel documentation.
Signed-off-by: Martin Waitz <tali@admingilde.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Dave Jones [Mon, 14 Nov 2005 00:08:09 +0000 (16:08 -0800)]
[PATCH] v4l: saa711x driver doesn't need segment.h
This breaks compilation on non-x86 architectures, and isn't even used.
Signed-off-by: Dave Jones <davej@redhat.com> Acked-by: Michael Krufky <mkrufky@m1k.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The following patch caused some duplicated code in cx88-dvb.c:
[PATCH] v4l: 634: implemented tuner set standby on cx88 init
The cx88-dvb.c portion of this patch was already applied
in an earlier patch, entitled:
[PATCH] v4l: fixup on cx88_dvb for Dvico HDTV5 Gold
I love quilt and all, but AFAIK, no tool is 100% perfect for catching
oversights like this.
The non-overlapping portions of each of these patches are still needed, and
must not be discarded, so rather than reverting old patches, please just apply
this fixup patch to remove the duplicated code.
Signed-off-by: Michael Krufky <mkrufky@m1k.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
[PATCH] v4l: (951) Make saa7134-oss as a stand-alone module
- saa7134-oss is now a standalone module as well
- remaining DMA sound code has been removed from core the module
- Lots of small cleanups and variable renames to get more consistency
between the OSS and ALSA drivers
- Fixed saa7134-alsa spinlock bug
- Added missing #include in saa7134-oss
[PATCH] v4l: (935) Moved common IR stuff to ir-common.c
- The pinnacle handler & remote are common to saa7134 PCI boards and em28xx
USB boards, so the keymap was moved to ir-common and the keyhandler is back
to ir-kbd-i2c
- request_module("ir-kbd-i2c") is no longer necessary at saa7134-core since
saa7134.ko now depends on ir-kbd-i2c.ko to get the keyhandler
George Anzinger [Mon, 14 Nov 2005 00:07:44 +0000 (16:07 -0800)]
[PATCH] timespec: normalize off by one errors
It would appear that the timespec normalize code has an off by one error.
Found in three places. Thanks to Ben for spotting.
Signed-off-by: George Anzinger<george@mvista.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Kylene Jo Hall [Mon, 14 Nov 2005 00:07:41 +0000 (16:07 -0800)]
[PATCH] tpm: updates for new hardware
This is the patch to support TPMs on power ppc hardware. It has been
reworked as requested to remove the need for messing with the io page mask
by just using ioremap.
Signed-off-by: Kylene Hall <kjhall@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Kylene Jo Hall [Mon, 14 Nov 2005 00:07:41 +0000 (16:07 -0800)]
[PATCH] tpm: necessary PPC64 function exports
Some work is needed in the tpm device driver to discover the TPM out of the
device tree rather than based on set address on Power PPC. This patch
exports a couple of functions for the parsing.
Signed-off-by: Kylene Hall <kjhall@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Zach Brown [Mon, 14 Nov 2005 00:07:35 +0000 (16:07 -0800)]
[PATCH] aio: don't ref kioctx after decref in put_ioctx
put_ioctx's refcount debugging was doing an atomic_read after dropping its
reference when it wasn't the last ref, leaving a tiny race for another freeing
thread to sneak into. This shifts the debugging before the ops, uses BUG_ON,
and reformats the defines a little. Sadly, moving to inlines increased the
code size but this change decreases the code size by a whole 9 bytes :)
Signed-off-by: Zach Brown <zach.brown@oracle.com> Cc: Benjamin LaHaise <bcrl@kvack.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Zach Brown [Mon, 14 Nov 2005 00:07:33 +0000 (16:07 -0800)]
[PATCH] aio: remove kioctx from mm_struct
Sync iocbs have a life cycle that don't need a kioctx. Their retrying, if
any, is done in the context of their owner who has allocated them on the
stack.
The sole user of a sync iocb's ctx reference was aio_complete() checking for
an elevated iocb ref count that could never happen. No path which grabs an
iocb ref has access to sync iocbs.
If we were to implement sync iocb cancelation it would be done by the owner of
the iocb using its on-stack reference.
Removing this chunk from aio_complete allows us to remove the entire kioctx
instance from mm_struct, reducing its size by a third. On a i386 testing box
the slab size went from 768 to 504 bytes and from 5 to 8 per page.
Signed-off-by: Zach Brown <zach.brown@oracle.com> Acked-by: Benjamin LaHaise <bcrl@kvack.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Kirill Korotaev [Mon, 14 Nov 2005 00:07:30 +0000 (16:07 -0800)]
[PATCH] stop_machine() vs. synchronous IPI send deadlock
This fixes deadlock of stop_machine() vs. synchronous IPI send. The
problem is that stop_machine() disables interrupts before disabling
preemption on other CPUs. So if another CPU is preempted and then calls
something like flush_tlb_all() it will deadlock with CPU doing
stop_machine() and which can't process IPI due to disabled IRQs.
I changed stop_machine() to do the same things exactly as it does on other
CPUs, i.e. it should disable preemption first on _all_ CPUs including
itself and only after that disable IRQs.
Signed-off-by: Kirill Korotaev <dev@sw.ru> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: "Andrey Savochkin" <saw@sawoct.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Nick Piggin [Mon, 14 Nov 2005 00:07:25 +0000 (16:07 -0800)]
[PATCH] atomic: inc_not_zero
Introduce an atomic_inc_not_zero operation. Make this a special case of
atomic_add_unless because lockless pagecache actually wants
atomic_inc_not_negativeone due to its offset refcount.
Signed-off-by: Nick Piggin <npiggin@suse.de> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Ingo Molnar [Mon, 14 Nov 2005 00:07:22 +0000 (16:07 -0800)]
[PATCH] rcutorture: renice to low priority
Make the box usable for interactive work when running the RCU torture test,
by renicing the RCU torture-test threads to +19 by default. Kthreads run
at nice -5 by default.
Signed-off-by: Ingo Molnar <mingo@elte.hu> Acked-by: Paul E. McKenney" <paulmck@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Paul Mundt [Mon, 14 Nov 2005 00:07:21 +0000 (16:07 -0800)]
[PATCH] Shut up per_cpu_ptr() on UP
Currently per_cpu_ptr() doesn't really do anything with 'cpu' in the UP
case. This is problematic in the cases where this is the only place the
variable is referenced:
CC kernel/workqueue.o
kernel/workqueue.c: In function `current_is_keventd':
kernel/workqueue.c:460: warning: unused variable `cpu'
Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Heiko Carstens [Mon, 14 Nov 2005 00:07:14 +0000 (16:07 -0800)]
[PATCH] signal handling: revert sigkill priority fix
This patch reverts commit c33880aaddbbab1ccf36f4457ed1090621f2e39a since
it's not needed anymore. As pointed out by Roland McGrath the real fix
is to deliver all signals before returning to user space.
See http://www.ussg.iu.edu/hypermail/linux/kernel/0509.2/0683.html
A fix for s390 has been merged.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Acked-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The access_ok_tt() macro is bogus, in that a read access is unconditionally
considered valid.
I couldn't find in SCM logs the introduction of this check, but I went back to
2.4.20-1um and the definition was the same.
Possibly this was done to avoid problems with missing set_fs() calls, but
there can't be any I think because they would fail with SKAS mode.
TT-specific code is still to check.
Also, this patch joins common code together, and makes the "address range
wrapping" check happen for all cases, rather than for only some.
This may, possibly, be reoptimized at some time, but the current code doesn't
seem clever, just confused.
* Important: I've also had to change references to access_ok_{tt,skas} back to
access_ok - the kernel wasn't that happy otherwise.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Acked-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
[PATCH] uml console channels: fix the API of console_write
Since the 4th param is unused, remove it altogether.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Acked-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
We were using a long series of (stupid) wrappers which all call
generic_console_write(). Since the wrappers only change the 4th param, which
is unused by the called proc, remove them and call generic_console_write()
directly.
If needed at any time in the future to reintroduce this stuff, the member
could be moved to a generic struct, to avoid this duplicated handling.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Acked-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
printk clears the host errno (I verified this in debugging and it's reasonable
enough, given that it ends via a write call on some fd, especially since
printk() goes on /dev/tty0 which is often the host stdout). So save errno
earlier. There's no reason to change the printk calls to use -err rather than
errno - the assignment can't clear errno.
And in the first failure path, we used to return 0 too (and this time more
clearly), which is totally wrong. 0 is a success fd, which is then registered
and gives a "registering fd twice" warning.
Finally, fix up some whitespace.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Acked-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
[PATCH] uml: fixups for "reuse i386 cpu-specific tuning"
A few fixups - show the new submenu only for x86 subarchitecture (it does not
make sense to show it for x86_64 users) and remove X86_CMPXCHG, which is now a
duplicate of Kconfig.i386, even though Kconfig doesn't complain (we also miss
the dependency on !M386 CPU).
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Acked-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Remove a stone-age comment (UM *does* have a MMU, i.e. the host), and fix a
dependency (introduced in commit 02edeb586ae4cdd17778923674700edb732a4741) to
do what was intended.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Acked-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
[PATCH] uml: remove bogus WARN_ON, triggerable harmlessly on a page fault race
The below warning was added in place of pte_mkyoung(); if (is_write)
pte_mkdirty();
In fact, if the PTE is not marked young/dirty, our dirty/accessed bit
emulation would cause the TLB permission not to be changed, and so we'd loop,
and given we don't support preemption yet, we'd busy-hang here.
However, I've seen this warning trigger without crashes during a loop of
concurrent kernel builds, at random times (i.e. like a race condition), and I
realized that two concurrent faults on the same page, one on read and one on
write, can trigger it. The read fault gets serviced and the PTE gets marked
writable but clean (it's possible on a shared-writable mapping), while the
generic code sees the PTE was already installed and returns without action. In
this case, we'll see another fault and service it normally.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Acked-by: Jeff Dike <jdike@addtoit.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
[PATCH] Kbuild: index asm-$(SUBARCH) headers for UML
In Uml, many definitions are borrowed from underlying subarch headers (with
#include <asm/arch/stuff.h>). And it has become annoying to keep switching
tag files all time, so by default index the underlying subarch headers too.
Btw, it adds negligible space to the tags file (less than 1M surely, IIRC it
was around 500k over 40M).
Finally, preserve the ALLSOURCE_ARCHS command line option (I hope) - if it is
set, it is used for headers too as before. But check my construct please, I
didn't test this.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Acked-by: Jeff Dike <jdike@addtoit.com> Cc: Sam Ravnborg <sam@ravnborg.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Roman Zippel [Mon, 14 Nov 2005 00:06:59 +0000 (16:06 -0800)]
[PATCH] m68k: convert thread flags to use bit fields
Remove task_work structure, use the standard thread flags functions and use
shifts in entry.S to test the thread flags. Add a few local labels to entry.S
to allow gas to generate short jumps.
Finally it changes a number of inline functions in thread_info.h to macros to
delay the current_thread_info() usage, which requires on m68k a structure
(task_struct) not yet defined at this point.
Signed-off-by: Roman Zippel <zippel@linux-m68k.org> Cc: Al Viro <viro@parcelfarce.linux.theplanet.co.uk> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Al Viro [Mon, 14 Nov 2005 00:06:57 +0000 (16:06 -0800)]
[PATCH] m68k: thread_info header cleanup
a) in smp_lock.h #include of sched.h and spinlock.h moved under #ifdef
CONFIG_LOCK_KERNEL.
b) interrupt.h now explicitly pulls sched.h (not via smp_lock.h from
hardirq.h as it used to)
c) in three more places we need changes to compensate for (a) - one place
in arch/sparc needs string.h now, hardirq.h needs forward declaration of
task_struct and preempt.h needs direct include of thread_info.h.
d) thread_info-related helpers in sched.h and thread_info.h put under
ifndef __HAVE_THREAD_FUNCTIONS. Obviously safe.
Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk> Signed-off-by: Roman Zippel <zippel@linux-m68k.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Al Viro [Mon, 14 Nov 2005 00:06:56 +0000 (16:06 -0800)]
[PATCH] m68k: introduce setup_thread_stack() and end_of_stack()
encapsulates the rest of arch-dependent operations with thread_info access.
Two new helpers - setup_thread_stack() and end_of_stack(). For normal case
the former consists of copying thread_info of parent to new thread_info and
the latter returns pointer immediately past the end of thread_info.
Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk> Signed-off-by: Roman Zippel <zippel@linux-m68k.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Al Viro [Mon, 14 Nov 2005 00:06:55 +0000 (16:06 -0800)]
[PATCH] m68k: introduce task_thread_info
new helper - task_thread_info(task). On platforms that have thread_info
allocated separately (i.e. in default case) it simply returns
task->thread_info. m68k wants (and for good reasons) to embed its thread_info
into task_struct. So it will (in later patch) have task_thread_info() of its
own. For now we just add a macro for generic case and convert existing
instances of its body in core kernel to uses of new macro. Obviously safe -
all normal architectures get the same preprocessor output they used to get.
Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk> Signed-off-by: Roman Zippel <zippel@linux-m68k.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The cpu_khz global is not initialized and remains 0 if you boot with
clock=pit, even if the processor does have a TSC. This may have bad
ramifications since the variable is used in various places scattered around
the kernel, though I didn't check them all to see if they can tolerate cpu_khz
= 0. You can observe the problem by doing "cat /proc/cpuinfo"; the cpu MHz
line says 0.000.
The fix is trivial; call init_cpu_khz() from init_pit(), just as it's called
from the timers/timer_foo.c:init_foo() for other values of foo.
Cc: john stultz <johnstul@us.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Adrian Bunk [Mon, 14 Nov 2005 00:06:51 +0000 (16:06 -0800)]
[PATCH] arch/i386/mm/init.c: small cleanups
This patch contains the following cleanups:
- make a needlessly global function static
- every file should include the headers containing the prototypes for
it's global functions
Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Andrew Morton [Mon, 14 Nov 2005 00:06:50 +0000 (16:06 -0800)]
[PATCH] powerpc-xmon-build-fix
arch/powerpc/xmon/xmon.c:525: error: syntax error before "xmon_irq"
arch/powerpc/xmon/xmon.c:526: warning: return type defaults to `int'
arch/powerpc/xmon/xmon.c: In function `xmon_irq':
arch/powerpc/xmon/xmon.c:532: error: `IRQ_HANDLED' undeclared (first use in this function)
arch/powerpc/xmon/xmon.c:532: error: (Each undeclared identifier is reported only once
arch/powerpc/xmon/xmon.c:532: error: for each function it appears in.)
Cc: Paul Mackerras <paulus@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Kumar Gala [Mon, 14 Nov 2005 00:06:49 +0000 (16:06 -0800)]
[PATCH] ppc32: Add support for handling PCI interrupts on MPC834x PCI expansion card
The MPC8349 PIBs system has a expansion board with 6 PCI slots. We needed
to update the IDSEL interrupt mapping for it to work properly. However,
only PCI1 is supported as the first revision of this expansion board
doesn't function properly for PCI2. For the time being we have zero'd out
the entries for the IDSELs related to PCI2. When a functioning expansion
board exists we can fix the table.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The slab allocator never uses alloc_pages since kmem_getpages() is always
called with a valid nodeid. Remove the branch and the code from
kmem_getpages()
Signed-off-by: Christoph Lameter <clameter@sgi.com> Cc: Manfred Spraul <manfred@colorfullife.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Nick Piggin [Mon, 14 Nov 2005 00:06:45 +0000 (16:06 -0800)]
[PATCH] mm: highmem watermarks
The pages_high - pages_low and pages_low - pages_min deltas are the asynch
reclaim watermarks. As such, the should be in the same ratios as any other
zone for highmem zones. It is the pages_min - 0 delta which is the
PF_MEMALLOC reserve, and this is the region that isn't very useful for
highmem.
This patch ensures highmem systems have similar characteristics as non highmem
ones with the same amount of memory, and also that highmem zones get similar
reclaim pressures to other zones.
Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Paul Jackson [Mon, 14 Nov 2005 00:06:44 +0000 (16:06 -0800)]
[PATCH] mm: gfp_noreclaim cleanup
Remove last remnant of the defunct early reclaim page logic, the no longer
used __GFP_NORECLAIM flag bit.
Signed-off-by: Paul Jackson <pj@sgi.com> Acked-by: Martin Hicks <mort@bork.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Robin Holt [Mon, 14 Nov 2005 00:06:42 +0000 (16:06 -0800)]
[PATCH] mm: ZAP_BLOCK causes redundant work
The address based work estimate for unmapping (for lockbreak) is and always
was horribly inefficient for sparse mappings. The problem is most simply
explained with an example:
If we find a pgd is clear, we still have to call into unmap_page_range
PGDIR_SIZE / ZAP_BLOCK_SIZE times, each time checking the clear pgd, in
order to progress the working address to the next pgd.
The fundamental way to solve the problem is to keep track of the end
address we've processed and pass it back to the higher layers.
From: Nick Piggin <npiggin@suse.de>
Modification to completely get away from address based work estimate
and instead use an abstract count, with a very small cost for empty
entries as opposed to present pages.
On 2.6.14-git2, ppc64, and CONFIG_PREEMPT=y, mapping and unmapping 1TB
of virtual address space takes 1.69s; with the following patch applied,
this operation can be done 1000 times in less than 0.01s
From: Andrew Morton <akpm@osdl.org>
With CONFIG_HUTETLB_PAGE=n:
mm/memory.c: In function `unmap_vmas':
mm/memory.c:779: warning: division by zero