POWER8 handles unaligned accesses in little endian mode, but commit
0b5e6661ac ("powerpc: Don't set HAVE_EFFICIENT_UNALIGNED_ACCESS on
little endian builds") disabled it for all.
The issue with unaligned little endian accesses is specific to POWER7,
so update the Kconfig check to match. Using the stat() testcase from
commit a75c380c71 ("powerpc: Enable DCACHE_WORD_ACCESS on ppc64le"),
performance improves 15% on POWER8.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
I see quite a lot of static branch mispredictions on a simple
web serving workload. The issue is in __atomic_add_unless(), called
from _atomic_dec_and_lock(). There is no obvious common case, so it
is better to let the hardware predict the branch.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
During context switch, switch_mm() sets our current CPU in mm_cpumask.
We can avoid this atomic sequence in most cases by checking before
setting the bit.
Testing on a POWER8 using our context switch microbenchmark:
tools/testing/selftests/powerpc/benchmarks/context_switch \
--process --no-fp --no-altivec --no-vector
Performance improves 2%.
Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Balbir Singh <bsingharora@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
No real need for this to be pr_warn(), reduce it to pr_info().
Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
We are starting to see i40e adapters in recent machines, so enable
it in our configs.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Change a few devices and filesystems that are seldom used any more
from built in to modules. This reduces our vmlinux about 500kB.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
When we issue a system reset, every CPU in the box prints an Oops,
including a backtrace. Each of these can be quite large (over 4kB)
and we may end up wrapping the ring buffer and losing important
information.
Bump the base size from 128kB to 256kB and the per CPU size from
4kB to 8kB.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
We see big improvements with the VMX crypto functions (often 10x or more),
so enable it as a module.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Align the hot loops in our assembly implementation of memset()
and backwards_memcpy().
backwards_memcpy() is called from tcp_v4_rcv(), so we might
want to optimise this a little more.
Signed-off-by: Anton Blanchard <anton@samba.org>
Reviewed-by: Nick Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This was not done before the big patches because I only noticed
them afterwards. It has become much easier to see which handlers
are branched to from which exception vectors now, and to see
exactly what vector space is being used for what.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Simple substitution. This is possible now that both parts of the OOL
initial handler get linked into their correct location.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
This is not an exception handler as such, it's called from
local_irq_enable(), not exception entry.
Also clean up some now redundant comments at the end of the
consolidation series.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Use assembler sections of fixed size and location to arrange the 64-bit
Book3S exception vector code (64-bit Book3E also uses it in head_64.S
for 0x0..0x100).
This allows better flexibility in arranging exception code and hiding
unimportant details behind macros.
Gas sections can be a bit painful to use this way, mainly because the
assembler does not know where they will be finally linked. Taking
absolute addresses requires a bit of trickery for example, but it can
be hidden behind macros for the most part.
Generated code is mostly the same except locations, offsets, alignments.
The "+ 0x2" is only required for the trap number / kvm exit number,
which gets loaded as a constant into a register.
Previously, code also used + 0x2 for label names, but we changed to
using "H" to distinguish HV case for that. Remove the last vestiges
of that.
__after_prom_start is taking absolute address of a label in another
fixed section. Newer toolchains seemed to compile this okay, but older
ones do not. FIXED_SYMBOL_ABS_ADDR is more foolproof, it just takes an
additional line to define.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
With a subsequent patch to put text into different sections,
(_end - _stext) can no longer be computed at link time to determine
the end of the copy. Instead, calculate it at runtime with
(copy_to_here - _stext) + (_end - copy_to_here).
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Move exception handler alignment directives into the head-64.h macros,
beause they will no longer work in-place after the next patch. This
slightly changes functions that have alignments applied and therefore
code generation, which is why it was not done initially (see earlier
patch).
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
