Madison cpus support 64 TR registers. Increase IA64_TR_ALLOC_MAX
to 64. Also fixup the messages that get printed when this limit
is exceeded. Repeating for every cpu is too noisy.
Signed-off-by: Tony Luck <tony.luck@intel.com>
If "max_purges" from PAL is 0, it actually means 1.
However it was not handled later when a hot-added cpu pass the
max_purges from PAL. This makes systems easy to go BUG_ON().
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
The patch defines kernel parameter "nptcg=". The parameter overrides max number
of concurrent global TLB purges which is reported from either PAL_VM_SUMMARY or
SAL PALO.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
According to SDM2.2, Itanium supports multiple outstanding ptc.g instructions.
But current kernel function ia64_global_tlb_purge() uses a spinlock to serialize
ptc.g instructions issued by multiple processors. This serialization might have
scalability issue on a big SMP machine where many processors could purge TLB
in parallel.
The patch fixes this problem by issuing multiple ptc.g instructions in
ia64_global_tlb_purge(). It also adds support for the "PALO" table to get
a platform view of the max number of outstanding ptc.g instructions (which
may be different from the processor view found from PAL_VM_SUMMARY).
PALO specification can be found at: http://www.dig64.org/home/DIG64_PALO_R1_0.pdf
spinaphore implementation by Matthew Wilcox.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Dynamic TR resource should be managed in the uniform way.
Add two interfaces for kernel:
ia64_itr_entry: Allocate a (pair of) TR for caller.
ia64_ptr_entry: Purge a (pair of ) TR by caller.
Signed-off-by: Xiantao Zhang <xiantao.zhang@intel.com>
Signed-off-by: Anthony Xu <anthony.xu@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Improve performance of memory allocations on ia64 by avoiding a global TLB
purge to purge a single page from the file cache. This happens whenever we
evict a page from the buffer cache to make room for some other allocation.
Test case: Run 'find /usr -type f | xargs cat > /dev/null' in the
background to fill the buffer cache, then run something that uses memory,
e.g. 'gmake -j50 install'. Instrumentation showed that the number of
global TLB purges went from a few millions down to about 170 over a 12
hours run of the above.
The performance impact is particularly noticeable under virtualization,
because a virtual TLB is generally both larger and slower to purge than
a physical one.
Signed-off-by: Christophe de Dinechin <ddd@hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Tell GCC to stop spewing out unnecessary warnings for unused variables
passed to functions as pointers for ia64 files.
Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
It was reported from a field customer that global spin lock ptcg_lock
is giving a lot of grief on munmap performance running on a large numa
machine. What appears to be a problem coming from flush_tlb_range(),
which currently unconditionally calls platform_global_tlb_purge().
For some of the numa machines in existence today, this function is
mapped into ia64_global_tlb_purge(), which holds ptcg_lock spin lock
while executing ptc.ga instruction.
Here is a patch that attempt to avoid global tlb purge whenever
possible. It will use local tlb purge as much as possible. Though the
conditions to use local tlb purge is pretty restrictive. One of the
side effect of having flush tlb range instruction on ia64 is that
kernel don't get a chance to clear out cpu_vm_mask. On ia64, this mask
is sticky and it will accumulate if process bounces around. Thus
diminishing the possible use of ptc.l. Thoughts?
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Acked-by: Jack Steiner <steiner@sgi.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
I originally thought this was an bug only in the SN code, but I think I
also see a hole in the generic IA64 tlb code. (Separate patch was sent
for the SN problem).
It looks like there is a bug in the TLB flushing code. During context switch,
kernel threads (kswapd, for example) inherit the mm of the task that was
previously running on the cpu. Normally, this is ok because the previous context
is still loaded into the RR registers. However, if the owner of the mm
migrates to another cpu, changes it's context number, and references a
page before kswapd issues a tlb_purge for that same page, the purge will be
done with a stale context number (& RR registers).
Signed-off-by: Tony Luck <tony.luck@intel.com>
wrap_mmu_context(), delayed_tlb_flush(), get_mmu_context() all
have an extra { } block which cause one extra indentation.
get_mmu_context() is particularly bad with 5 indentations to
the most inner "if". It finally gets on my nerve that I can't
keep the code within 80 columns. Remove the extra { } block
and while I'm at it, reformat all the comments to 80-column
friendly. No functional change at all with this patch.
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Corrects the very inefficent method of finding free context_ids in
get_mmu_context(). Instead of walking the task_list of all processes,
2 bitmaps are used to efficently store and lookup state, inuse and
needs flushing. The entire rid address space is now used before calling
wrap_mmu_context and global tlb flushing.
Special thanks to Ken and Rohit for their review and modifications in
using a bit flushmap.
Signed-off-by: Peter Keilty <peter.keilty@hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
There was one small but very significant change in the previous patch:
mprotect's flush_tlb_range fell outside the page_table_lock: as it is in 2.4,
but that doesn't prove it safe in 2.6.
On some architectures flush_tlb_range comes to the same as flush_tlb_mm, which
has always been called from outside page_table_lock in dup_mmap, and is so
proved safe. Others required a deeper audit: I could find no reliance on
page_table_lock in any; but in ia64 and parisc found some code which looks a
bit as if it might want preemption disabled. That won't do any actual harm,
so pending a decision from the maintainers, disable preemption there.
Remove comments on page_table_lock from flush_tlb_mm, flush_tlb_range and
flush_tlb_page entries in cachetlb.txt: they were rather misleading (what
generic code does is different from what usually happens), the rules are now
changing, and it's not yet clear where we'll end up (will the generic
tlb_flush_mmu happen always under lock? never under lock? or sometimes under
and sometimes not?).
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
flush_tlb_all() can be a scaling issue on large SGI Altix systems
since it uses the global call_lock and always executes on all cpus.
When a process enters flush_tlb_range() to purge TLBs for another
process, it is possible to avoid flush_tlb_all() and instead allow
sn2_global_tlb_purge() to purge TLBs only where necessary.
This patch modifies flush_tlb_range() so that this case can be handled
by platform TLB purge functions and updates ia64_global_tlb_purge()
accordingly. sn2_global_tlb_purge() now calculates the region register
value from the mm argument introduced with this patch.
Signed-off-by: Dean Roe <roe@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
In arch/ia64 change the explicit use of for-loops and NR_CPUS into the
general for_each_cpu() or for_each_online_cpu() constructs, as
appropriate. This widens the scope of potential future optimizations
of the general constructs, as well as takes advantage of the existing
optimizations of first_cpu() and next_cpu().
Signed-off-by: John Hawkes <hawkes@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
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!
