This is the core of the (much simplified) early reclaim. The goal of this
patch is to reclaim some easily-freed pages from a zone before falling back
onto another zone.
One of the major uses of this is NUMA machines. With the default allocator
behavior the allocator would look for memory in another zone, which might be
off-node, before trying to reclaim from the current zone.
This adds a zone tuneable to enable early zone reclaim. It is selected on a
per-zone basis and is turned on/off via syscall.
Adding some extra throttling on the reclaim was also required (patch
4/4). Without the machine would grind to a crawl when doing a "make -j"
kernel build. Even with this patch the System Time is higher on
average, but it seems tolerable. Here are some numbers for kernbench
runs on a 2-node, 4cpu, 8Gig RAM Altix in the "make -j" run:
wall user sys %cpu ctx sw. sleeps
---- ---- --- ---- ------ ------
No patch 1009 1384 847 258 298170 504402
w/patch, no reclaim 880 1376 667 288 254064 396745
w/patch & reclaim 1079 1385 926 252 291625 548873
These numbers are the average of 2 runs of 3 "make -j" runs done right
after system boot. Run-to-run variability for "make -j" is huge, so
these numbers aren't terribly useful except to seee that with reclaim
the benchmark still finishes in a reasonable amount of time.
I also looked at the NUMA hit/miss stats for the "make -j" runs and the
reclaim doesn't make any difference when the machine is thrashing away.
Doing a "make -j8" on a single node that is filled with page cache pages
takes 700 seconds with reclaim turned on and 735 seconds without reclaim
(due to remote memory accesses).
The simple zone_reclaim syscall program is at
http://www.bork.org/~mort/sgi/zone_reclaim.c
Signed-off-by: Martin Hicks <mort@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
current->blocked will be set to the value of current->thread_info->flags if the
cmpxchg to update thread_info->flags fails. For performance reasons the store into
current->blocked was placed in the cmpxchg loop. However, the cmpxchg overwrites the
register holding the value to be stored. In the rare case of a retry the value of
thread_info->flags will be written into current->blocked.
The fix is to use another register so that the register containing the current->blocked
value is not overwritten.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
There've been reports of problems with CONFIG_PREEMPT=y and the high
floating point partition. This is caused by the possibility of preemption
and rescheduling on a different processor while saving or restioirng the
high partition.
The only places where the FPU state is touched are in ptrace, in
switch_to(), and where handling a floating-point exception. In switch_to()
preemption is off. So it's only in trap.c and ptrace.c that we need to
prevent preemption.
Here is a patch that adds commentary to make the conditions clear, and adds
appropriate preempt_{en,dis}able() calls to make it so. In trap.c I use
preempt_enable_no_resched(), as we're about to return to user space where
the preemption flag will be checked anyway.
Signed-off-by: Peter Chubb <peterc@gelato.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
break.b does not store the break number in cr.iim, instead it stores 0,
which makes all break.b instructions look like BUG(). Extract the
break number from the instruction itself.
Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Christian Hildner pointed out that the comment did not match what the
code does in cpu_init() when we set up the default control register.
Patch based on suggestions from Ken Chen.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Some bits of the kernel assume that gp always points to valid memory,
in particular PHYSICAL_MODE_ENTER() assumes that both gp and sp are
valid virtual addresses with associated physical pages. The IA64
module loader puts gp well past the end of the module, with no physical
backing. Offsets on gp are still valid, but physical mode addressing
breaks for modules. Ensure that gp always falls within the module
body. Also ensure that gp is 8 byte aligned.
Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Fix a bug in which shub_1_1_found is not being properly initialized or set,
resulting in the improper setting of sn_hub_info->shub_1_1_found.
Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
This gets rid of an unused variable `error' in sys_ia32.c:sys32_epoll_wait()
Getting rid of this one makes parsing the output of the kernecomp
autobuild easier --- searching for `Error' to find a problem kept
hitting this one, even though it's only a warning.
Signed-off-by: Tony Luck <tony.luck@intel.com>
The attached patch cleans up a compilation warning when ACPI
is turned off (i.e., when compiling for the Ski simulator).
Signed-off-by: Tony Luck <tony.luck@intel.com>
In IA64 kernel, sys_mmap calls do_mmap2 and do_mmap2 returns addr if
len=0, which means the mmap sys call succeeds.
Posix.1 says:
The mmap() function shall fail if:
[EINVAL] The value of len is zero.
Here is a patch to fix it.
Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
Acked-by: David Mosberger <davidm@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
I applied the penultimate version of the perfmon patch, which didn't have
the initialization of the new spinlock that was added.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Patch from Charles Spirakis
Some linux customers want to optimize their applications on the latest
hardware but are not yet willing to upgrade to the latest kernel. This
patch provides a way to plug in an alternate, basic, and GPL'ed PMU
subsystem to help with their monitoring needs or for specialty work. It
can also be used in case of serious unexpected bugs in perfmon. Mutual
exclusion between the two subsystems is guaranteed, hence no conflict
can arise from both subsystem being present.
Acked-by: Stephane Eranian <eranian@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
The 2.6 kernel has CPE error thresholding.
This patch lets SAL know of this error handling feature.
The changes are SN specific.
Signed-off-by: Russ Anderson (rja@sgi.com)
Signed-off-by: Tony Luck <tony.luck@intel.com>
acpi_request_vector() is called in ia64_mca_init() to get the cpe_vector.
The problem is that acpi_request_vector() looks in platform_intr_list[] to
get the vector, but platform_intr_list[] is not initialized with a valid
vector until later (in sn_setup()). Without a valid vector the code
defaults to polling mode.
This patch moves the call to acpi_request_vector() from ia64_mca_init()
to ia64_mca_late_init(), which is after platform_intr_list[] is initialized.
Signed-off-by: Russ Anderson (rja@sgi.com)
Signed-off-by: Tony Luck <tony.luck@intel.com>
convert_to_non_syscall() has the same problem that unwind_to_user()
used to have. Fix it likewise.
Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
These days <linux/ioctl32.h> handles everything, no need for an asm
header on just two architectures.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Some time ago, GAS was fixed to bring the .spillpsp directive in line
with the Intel assembler manual (there was some disagreement as to
whether or not there is a built-in 16-byte offset). Unfortunately,
there are two places in the kernel where this directive is used in
handwritten assembly files and those of course relied on the "buggy"
behavior. As a result, when using a "fixed" assembler, the kernel
picks up the UNaT bits from the wrong place (off by 16) and randomly
sets NaT bits on the scratch registers. This can be noticed easily by
looking at a coredump and finding various scratch registers with
unexpected NaT values. The patch below fixes this by using the
.spillsp directive instead, which works correctly no matter what
assembler is in use.
Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
I noticed this typo when trying to compile a kernel which had
CONFIG_HOTPLUG turned off. In that case, __devinit is no longer a
no-op and the compiler then detects a section-conflict. Fix by using
__devinitdata instead of __devinit.
Same patch also submitted by Darren Williams to fix compilation error
using sim_defconfig (which has CONFIG_HOTPLUG=n).
Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com>
Signed-off-by: Darren Williams <dsw@gelato.unsw.edu.au>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Without this patch, the stack is placed _below_ the current task
structure, which is risky at best.
Tony, I think this patch needs to go into 2.6.12, since it fixes a
real bug. Without it, INIT may case secondary errors, which would be
most unpleasant.
Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
While looking at code generated by gcc4.0 I noticed some functions still
had frame pointers, even after we stopped ppc64 from defining
CONFIG_FRAME_POINTER. It turns out kernel/Makefile hardwires
-fno-omit-frame-pointer on when compiling schedule.c.
Create CONFIG_SCHED_NO_NO_OMIT_FRAME_POINTER and define it on architectures
that dont require frame pointers in sched.c code.
(akpm: blame me for the name)
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Remove the p_nodepda and p_subnodepda pointers from the pda_s structure.
And then define a new per-cpu pointer to the nodepda and export it so
that it can be accessed by kernel modules.
Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
- pfm_context_load(): change return value from EINVAL to EBUSY
when context is already loaded.
- pfm_check_task_state(): pass test if context state is MASKED.
It is safe to give access on PFM_CTX_MASKED because the PMU
state (PMD) is stable and saved in software state.
This helps multiplexing programs such as the example given
in libpfm-3.1.
Signed-off-by: stephane eranian <eranian@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
The pmu_active test is based on the values of PSR.up. THIS IS THE PROBLEM as
it does not take into account the lazy restore logic which is as follow (simplified):
context switch out:
save PMDs
clear psr.up
release ownership
context switch in:
if (ctx->last_cpu == smp_processor_id() && ctx->cpu_activation == cpu_activation) {
set psr.up
return
}
restore PMD
restore PMC
ctx->last_cpu = smp_processor_id();
ctx->activation = ++cpu_activation;
set psr.up
The key here is that on context switch out, we clear psr.up and on context switch in
we check if nobody else used the PMU on that processor since last time we came. In
that case, we assume the PMD/PMC are ours and we simply reactivate.
The Caliper problem is that between the moment we context switch out and the moment we
come back, nobody effectively used the PMU BUT the processor went idle. Normally this
would have no incidence but PAL_HALT does alter the PMU registers. In default_idle(),
the test on psr.up is not strong enough to cover this case and we go into PAL which
trashed the PMU resgisters. When we come back we falsely assume that this is our state
yet it is corrupted. Very nasty indeed.
To avoid the problem it is necessary to forbid going to PAL_HALT as soon as perfmon
installs some valid state in the PMU registers. This happens with an application
attaches a context to a thread or CPU. It is not enough to check the psr/dcr bits.
Hence I propose the attached patch. It adds a callback in process.c to modify the
condition to enter PAL on idle. Basically, now it is conditional to pal_halt=1 AND
perfmon saying it is okay.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Correct a bug where tioca_dma_mapped() is putting tioca dma map structs
on the wrong list.
Signed-off-by: Mark Maule <maule@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
When SAL calls back into the OS, the OS code is running with preempt
disabled so it cannot call sleeping functions.
Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Jack Steiner uncovered some opportunities for improvement in
the MCA recovery code.
1) Set bsp to save registers on the kernel stack.
2) Disable interrupts while in the MCA recovery code.
3) Change the way the user process is killed, to avoid
a panic in schedule.
Testing shows that these changes make the recovery code much
more reliable with the 2.6.12 kernel.
Signed-off-by: Russ Anderson <rja@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Attached is a patch against David's audit.17 kernel that adds checks
for the TIF_SYSCALL_AUDIT thread flag to the ia64 system call and
signal handling code paths. The patch enables auditing of system
calls set up via fsys_bubble_down, as well as ensuring that
audit_syscall_exit() is called on return from sigreturn.
Neglecting to check for TIF_SYSCALL_AUDIT at these points results in
incorrect information in audit_context, causing frequent system panics
when system call auditing is enabled on an ia64 system.
I have tested this patch and have seen no problems with it.
[Original patch from Amy Griffis ported to current kernel by David Woodhouse]
From: Amy Griffis <amy.griffis@hp.com>
From: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Chris Wright <chrisw@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Andi noted that during normal runtime cpu_idle_map is bounced around a lot,
and occassionally at a higher frequency than the timer interrupt wakeup
which we normally exit pm_idle from. So switch to a percpu variable.
I didn't move things to the slow path because it would involve adding
scheduler code to wakeup the idle thread on the cpus we're waiting for.
Signed-off-by: Zwane Mwaikambo <zwane@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Tony Luck <tony.luck@intel.com>
The following patch fixes a bug in the SGI Altix sn_dma_flush code.
sn_dma_flush is broken in 2.6. The code isn't waiting for the DMA
data to be flushed out of the PIC ASIC. This patch is based off the
linux-ia64-test-2.6.12 tree
Signed-off-by: Mike Habeck <habeck@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
This patch simplifies a couple places where we search for _PXM
values in ACPI namespace. Thanks,
Signed-off-by: Alex Williamson <alex.williamson@hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
The following patch ensures that the correct error interrupt handling
routine is initialized. This patch is based on the 2.6.12 ia64 release tree.
Signed-off-by: Colin Ngam <cngam@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
This patch detects the existence of an uncached physical AMO address setup
by EFI's XPBOOT (SGI) and converts it to an uncached virtual AMO address.
Depends on a patch submitted on 23 March 2005 with the subject of:
[PATCH 2/3] SGI Altix cross partition functionality (2nd revision)
Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
This patch contains the cross partition pseudo-ethernet driver (XPNET)
functional support module.
Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
This patch contains the communication module (XPC) for cross partition
communication on a partitioned SGI Altix.
Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
cg-patch couldn't apply the patch to Makefile, and my dumb script
rushed on and ran cg-commit without this change.
Signed-off-by: Tony Luck <tony.luck@intel.com>
This patch contains the shim module (XP) which interfaces between the
communication module (XPC) and the functional support modules (like XPNET).
Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Another step in the effort to eliminate the SN pda structure.
This patch moves the cnodeid_to_nasid_table field out of the pda,
making it a standalone per-cpu data item, and exports it so it can
be accessed by kernel modules.
Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Here is a patch to enable the SGI tiocx bus driver to distingush between
FPGA-attached h/w and non-FPGA-attached h/w.
Signed-off-by: Bruce Losure <blosure@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Hi Tony,
This patch against ia64-test-2.6.12 fixes a bug where the tiocx code
was inadvertently un-doing some address modifications done in earlier
fixup code. This patch just removes the offending code.
Signed-off-by: Bruce Losure <blosure@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
This is a small patch to switch fluch_icache_range() to use fc.i
instead of fc. This would save time on processors which can establish
i-cache coherency without flushing the cache-line out to memory (not
that any current processors do). On existing processors, fc.i behaves
like fc. The only caveat is that very old assemblers may not know
about fc.i yet.
Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Patch below fixes 3 trivial typos which are caught by the new
assembler (v2.169.90). Please apply.
[Note: fix to memcpy that was also part of this patch was separately
applied from patches by H.J. and Andreas ... so the delta here only
has the other two fixes. -Tony]
Signed-off-by: David Mosberger-Tang <davidm@hpl.hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
The current ia64 assembler complains about mismatching .proc/.endp pairs.
(Same patch also sent by H.J. Lu)
Signed-off-by: Andreas Schwab <schwab@suse.de>
Signed-off-by: Tony Luck <tony.luck@intel.com>
