| Age | Commit message (Collapse) | Author |
|---|
|
Ingo recently introduced a great speedup for allocating new mmaps using the
free_area_cache pointer which boosts the specweb SSL benchmark by 4-5% and
causes huge performance increases in thread creation.
The downside of this patch is that it does lead to fragmentation in the
mmap-ed areas (visible via /proc/self/maps), such that some applications
that work fine under 2.4 kernels quickly run out of memory on any 2.6
kernel.
The problem is twofold:
1) the free_area_cache is used to continue a search for memory where
the last search ended. Before the change new areas were always
searched from the base address on.
So now new small areas are cluttering holes of all sizes
throughout the whole mmap-able region whereas before small holes
tended to close holes near the base leaving holes far from the base
large and available for larger requests.
2) the free_area_cache also is set to the location of the last
munmap-ed area so in scenarios where we allocate e.g. five regions of
1K each, then free regions 4 2 3 in this order the next request for 1K
will be placed in the position of the old region 3, whereas before we
appended it to the still active region 1, placing it at the location
of the old region 2. Before we had 1 free region of 2K, now we only
get two free regions of 1K -> fragmentation.
The patch addresses thes issues by introducing yet another cache descriptor
cached_hole_size that contains the largest known hole size below the
current free_area_cache. If a new request comes in the size is compared
against the cached_hole_size and if the request can be filled with a hole
below free_area_cache the search is started from the base instead.
The results look promising: Whereas 2.6.12-rc4 fragments quickly and my
(earlier posted) leakme.c test program terminates after 50000+ iterations
with 96 distinct and fragmented maps in /proc/self/maps it performs nicely
(as expected) with thread creation, Ingo's test_str02 with 20000 threads
requires 0.7s system time.
Taking out Ingo's patch (un-patch available per request) by basically
deleting all mentions of free_area_cache from the kernel and starting the
search for new memory always at the respective bases we observe: leakme
terminates successfully with 11 distinctive hardly fragmented areas in
/proc/self/maps but thread creating is gringdingly slow: 30+s(!) system
time for Ingo's test_str02 with 20000 threads.
Now - drumroll ;-) the appended patch works fine with leakme: it ends with
only 7 distinct areas in /proc/self/maps and also thread creation seems
sufficiently fast with 0.71s for 20000 threads.
Signed-off-by: Wolfgang Wander <wwc@rentec.com>
Credit-to: "Richard Purdie" <rpurdie@rpsys.net>
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Acked-by: Ingo Molnar <mingo@elte.hu> (partly)
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
A lot of the code in arch/*/mm/hugetlbpage.c is quite similar. This patch
attempts to consolidate a lot of the code across the arch's, putting the
combined version in mm/hugetlb.c. There are a couple of uglyish hacks in
order to covert all the hugepage archs, but the result is a very large
reduction in the total amount of code. It also means things like hugepage
lazy allocation could be implemented in one place, instead of six.
Tested, at least a little, on ppc64, i386 and x86_64.
Notes:
- this patch changes the meaning of set_huge_pte() to be more
analagous to set_pte()
- does SH4 need s special huge_ptep_get_and_clear()??
Acked-by: William Lee Irwin <wli@holomorphy.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
This patch implements a number of smp_processor_id() cleanup ideas that
Arjan van de Ven and I came up with.
The previous __smp_processor_id/_smp_processor_id/smp_processor_id API
spaghetti was hard to follow both on the implementational and on the
usage side.
Some of the complexity arose from picking wrong names, some of the
complexity comes from the fact that not all architectures defined
__smp_processor_id.
In the new code, there are two externally visible symbols:
- smp_processor_id(): debug variant.
- raw_smp_processor_id(): nondebug variant. Replaces all existing
uses of _smp_processor_id() and __smp_processor_id(). Defined
by every SMP architecture in include/asm-*/smp.h.
There is one new internal symbol, dependent on DEBUG_PREEMPT:
- debug_smp_processor_id(): internal debug variant, mapped to
smp_processor_id().
Also, i moved debug_smp_processor_id() from lib/kernel_lock.c into a new
lib/smp_processor_id.c file. All related comments got updated and/or
clarified.
I have build/boot tested the following 8 .config combinations on x86:
{SMP,UP} x {PREEMPT,!PREEMPT} x {DEBUG_PREEMPT,!DEBUG_PREEMPT}
I have also build/boot tested x64 on UP/PREEMPT/DEBUG_PREEMPT. (Other
architectures are untested, but should work just fine.)
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The initial peek read PIO of the match register is just a waste.
Just do the flush writes first, as that is more efficient.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Firstly, if the direction is TODEVICE, then dirty data in the
streaming cache is impossible so we can elide the flush-flag
synchronization in that case.
Next, the context allocator is broken. It is highly likely
that contexts get used multiple times for different dma
mappings, which confuses the strbuf flushing code and makes
it run inefficiently.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Older UltraSPARC-III chips have a P-Cache bug that makes us disable it
by default at boot time.
However, this does hurt performance substantially, particularly with
memcpy(), and the bug is _incredibly_ obscure. I have never seen it
triggered in practice, ever.
So provide a "-P" boot option that forces the P-Cache on. It taints
the kernel, so if it does trigger and cause some data corruption or
OOPS, we will find out in the logs that this option was on when it
happened.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The recent change to add a timeout to strbuf flushing had
a negative performance impact. The udelay()'s are too long,
and they were done in the wrong order wrt. the register read
checks. Fix both, and things are happy again.
There are more possible improvements in this area. In fact,
PCI streaming buffer flushing seems to be part of the bottleneck
in network receive performance on my SunBlade1000 box.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
If some hardware error occurs and the flush flag never updates,
we will hang forever in these routines. Add a timeout, and
print out a diagnostic if it is reached.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Currently sparc and sparc64's UP cpu_idle() checks current pid. This
is old time legacy. Now it's paranoia.
Signed-off-by: Coywolf Qi Hunt <coywolf@lovecn.org>
Acked-by: William Irwin <wli@holomorphy.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
warning: untested, but it there's not too much chance for screwups
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There is some race whereby IRQs get stuck, the IRQ status
is pending but no processor actually handles the IRQ vector
and thus the interrupt.
This is a temporary workaround.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
We would never advance the goal_cpu counter like we
should, so all IRQs would go to a single processor.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Convert most of the current code that uses _NSIG directly to instead use
valid_signal(). This avoids gcc -W warnings and off-by-one errors.
Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
void * __iomem foo is not a pointer to iomem - it's an iomem variable
containing void *. A pile of such guys in arch/sparc64/kernel/time.c,
drivers/sbus/char/rtc.c and include/asm-sparc64/mostek.h turned into
intended void __iomem *.
Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
CONFIG_HW_CONSOLE selects vt.c; without the stuff pulled by CONFIG_VT it
will not build. Normally we get both in drivers/char/Kconfig and there
HW_CONSOLE depends on VT. sparc64 does not pull drivers/char/Kconfig
and has that sutff in arch/sparc64/Kconfig instead. However, it forgets
to add the same dependency. As the result, turning VT off [which is
possible] will end up with broken build. For no good reason...
Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Provide support for drivers/char/rtc.c ioctls in the
Mostek rtc driver as well as the Sparc specific RTCGET
and RTCSET.
This allows userspace to be much less messy. Currently
util-linux and other spots jump through hoops trying
various ioctl variants until it hits the right one whatever
driver actually being used supports.
Eventually all of this should move over to the genrtc.c
driver, but not today...
While we are here, fix up the register types for sparse.
Thanks to Frans Pop for helping point out this issue.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Like Alpha, sparc64's struct stat was defined before we had the
nanosecond et al. fields added. So like Alpha I have to cons up a
struct stat64 to get this stuff. I'll work on the glibc bits soon.
Also, we were forgetting to fill in the nanosecond fields in the sparc
compat stat64 syscalls.
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
The compat routine to copy over this data structure was not
handling SI_POLL correctly, breaking various fcntl() variants
in compat tasks.
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
We were flushing the D-cache excessively for ptrace() processing
and this makes debugging threads so slow as to be totally unusable.
All process page accesses via ptrace() go via access_process_vm().
This routine, for each process page, uses get_user_pages(). That
in turn does a flush_dcache_page() on the child pages before we
copy in/out the ptrace request data.
Therefore, all we need to do after the data movement is:
1) Flush the D-cache pages if the kernel maps the page to a different
color than userspace does.
2) If we wrote to the page, we need to flush the I-cache on older cpus.
Previously we just flushed the entire cache at the end of a ptrace()
request, and that was beyond stupid.
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
SunOS aparently had this weird PTRACE_CONT semantic which
we copied. If the addr argument is something other than
1, it sets the process program counter to whatever that
value is.
This is different from every other Linux architecture, which
don't do anything with the addr and data args.
This difference in particular breaks the Linux native GDB support
for fork and vfork tracing on sparc and sparc64.
There is no interest in running SunOS binaries using this weird
PTRACE_CONT behavior, so just delete it so we behave like other
platforms do.
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
A couple message queue system call entries for compat tasks
were not using the necessary compat_sys_*() functions, causing
some glibc test cases to fail.
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
This case actually can get exercised a lot during an ELF
coredump of a process which contains a lot of non-COW'd
anonymous pages. GDB has this test case which in partiaular
creates near terabyte process full of ZERO_PAGEes. It takes
forever to just walk through the page tables because of
all of these spurious cache flushes on sparc64.
With this change it takes only a second or so.
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
|
