linux-toradex.git/arch/powerpc/kernel/vdso64, branch v4.2-rc8

powerpc/booke64: Use SPRG7 for VDSO

2014-03-20T00:57:14+00:00

Previously SPRG3 was marked for use by both VDSO and critical
interrupts (though critical interrupts were not fully implemented).

In commit 8b64a9dfb091f1eca8b7e58da82f1e7d1d5fe0ad ("powerpc/booke64:
Use SPRG0/3 scratch for bolted TLB miss & crit int"), Mihai Caraman
made an attempt to resolve this conflict by restoring the VDSO value
early in the critical interrupt, but this has some issues:

 - It's incompatible with EXCEPTION_COMMON which restores r13 from the
   by-then-overwritten scratch (this cost me some debugging time).
 - It forces critical exceptions to be a special case handled
   differently from even machine check and debug level exceptions.
 - It didn't occur to me that it was possible to make this work at all
   (by doing a final "ld r13, PACA_EXCRIT+EX_R13(r13)") until after
   I made (most of) this patch. :-)

It might be worth investigating using a load rather than SPRG on return
from all exceptions (except TLB misses where the scratch never leaves
the SPRG) -- it could save a few cycles.  Until then, let's stick with
SPRG for all exceptions.

Since we cannot use SPRG4-7 for scratch without corrupting the state of
a KVM guest, move VDSO to SPRG7 on book3e.  Since neither SPRG4-7 nor
critical interrupts exist on book3s, SPRG3 is still used for VDSO
there.

Signed-off-by: Scott Wood 
Cc: Mihai Caraman 
Cc: Anton Blanchard 
Cc: Paul Mackerras 
Cc: kvm-ppc@vger.kernel.org

powerpc: Use unstripped VDSO image for more accurate profiling data

2014-02-17T00:19:37+00:00

We are seeing a lot of hits in the VDSO that are not resolved by perf.
A while(1) gettimeofday() loop shows the issue:

27.64%  [vdso]  [.] 0x000000000000060c
22.57%  [vdso]  [.] 0x0000000000000628
16.88%  [vdso]  [.] 0x0000000000000610
12.39%  [vdso]  [.] __kernel_gettimeofday
 6.09%  [vdso]  [.] 0x00000000000005f8
 3.58%  test    [.] 00000037.plt_call.gettimeofday@@GLIBC_2.18
 2.94%  [vdso]  [.] __kernel_datapage_offset
 2.90%  test    [.] main

We are using a stripped VDSO image which means only symbols with
relocation info can be resolved. There isn't a lot of point to
stripping the VDSO, the debug info is only about 1kB:

4680 arch/powerpc/kernel/vdso64/vdso64.so
5815 arch/powerpc/kernel/vdso64/vdso64.so.dbg

By using the unstripped image, we can resolve all the symbols in the
VDSO and the perf profile data looks much better:

76.53%  [vdso]  [.] __do_get_tspec
12.20%  [vdso]  [.] __kernel_gettimeofday
 5.05%  [vdso]  [.] __get_datapage
 3.20%  test    [.] main
 2.92%  test    [.] 00000037.plt_call.gettimeofday@@GLIBC_2.18

Signed-off-by: Anton Blanchard 
Signed-off-by: Benjamin Herrenschmidt

powerpc: Delete non-required instances of include

2014-01-15T02:46:44+00:00

None of these files are actually using any __init type directives
and hence don't need to include .  Most are just a
left over from __devinit and __cpuinit removal, or simply due to
code getting copied from one driver to the next.

The one instance where we add an include for init.h covers off
a case where that file was implicitly getting it from another
header which itself didn't need it.

Signed-off-by: Paul Gortmaker 
Signed-off-by: Benjamin Herrenschmidt

powerpc: Wrong DWARF CFI in the kernel vdso for little-endian / ELFv2

2013-11-20T22:19:23+00:00

I've finally tracked down why my CR signal-unwind test case still
fails on little-endian.  The problem turned to be that the kernel
installs a signal trampoline in the vDSO, and provides a DWARF CFI
record for that trampoline.  This CFI describes the save location
for CR:

  rsave (70, 38*RSIZE + (RSIZE - CRSIZE))

which is correct for big-endian, but points to the wrong word on
little-endian.   This is wrong no matter which ABI.

In addition, for the ELFv2 ABI, we should not only provide a CFI
record for register 70 (cr2), but for all CR fields separately.
Strictly speaking, I guess this would mean providing two separate
vDSO images, one for ELFv1 processes and one for ELFv2 processes (or
maybe playing some tricks with conditional DWARF expressions).
However, having CFI records for the other CR fields in ELFv1 is not
actually wrong, they just will be ignored.   So it seems the simplest
fix would be just to always provide CFI for all the fields.

Signed-off-by: Ulrich Weigand 
Signed-off-by: Anton Blanchard 
Signed-off-by: Benjamin Herrenschmidt

powerpc: Add ability to build little endian kernels

2013-10-11T05:48:56+00:00

This patch allows the kbuild system to successfully compile a kernel
for the little endian PowerPC64 architecture. A subsequent patch
will add the CONFIG_CPU_LITTLE_ENDIAN kernel config option which
must be set to build such a kernel.

If cross compiling, CROSS_COMPILE must point to a suitable toolchain
(compiled for the powerpc64le-linux and powerpcle-linux targets).

Signed-off-by: Ian Munsie 
Signed-off-by: Anton Blanchard 
Signed-off-by: Benjamin Herrenschmidt

powerpc: Add VDSO version of time

2013-04-23T06:05:05+00:00

On 04/18/2013 07:38 PM, Anton Blanchard wrote:
> Since you are only reading one long you shouldn't need to check the
> update count and loop, you will always see a consistent value. The
> system call version of time() just does an unprotected load for example.

Fixed.

> With the above change and with Michael's comments covered (decent
> changelog entry and Signed-off-by):
>
> Acked-by: Anton Blanchard 

Thanks for the review, below the updated patch:

From: Adhemerval Zanella 

This patch implement the time syscall as vDSO. The performance speedups
are:

Baseline PPC32: 380 nsec
Baseline PPC64: 350 nsec
vdso PPC32:      20 nsec
vsdo PPC64:      20 nsec

Tested on 64 bit build with both 32 bit and 64 bit userland.

Acked-by: Anton Blanchard 
Signed-off-by: Adhemerval Zanella 
Signed-off-by: Benjamin Herrenschmidt

powerpc: Add VDSO version of getcpu

2012-07-11T04:18:40+00:00

We have a request for a fast method of getting CPU and NUMA node IDs
from userspace. This patch implements a getcpu VDSO function,
similar to x86.

Ben suggested we use SPRG3 which is userspace readable. SPRG3 can be
modified by a KVM guest, so we save the SPRG3 value in the paca and
restore it when transitioning from the guest to the host.

I have a glibc patch that implements sched_getcpu on top of this.
Testing on a POWER7:

baseline: 538 cycles
vdso:      30 cycles

Signed-off-by: Anton Blanchard 
Signed-off-by: Benjamin Herrenschmidt

Fix common misspellings

2011-03-31T14:26:23+00:00

Fixes generated by 'codespell' and manually reviewed.

Signed-off-by: Lucas De Marchi

powerpc/Makefiles: Change to new flag variables

2010-10-13T05:19:22+00:00

Replace EXTRA_CFLAGS with ccflags-y and EXTRA_AFLAGS with asflags-y.

Signed-off-by: matt mooney 
Signed-off-by: Benjamin Herrenschmidt

powerpc: Rework VDSO gettimeofday to prevent time going backwards

2010-07-09T01:26:16+00:00

Currently it is possible for userspace to see the result of
gettimeofday() going backwards by 1 microsecond, assuming that
userspace is using the gettimeofday() in the VDSO.  The VDSO
gettimeofday() algorithm computes the time in "xsecs", which are
units of 2^-20 seconds, or approximately 0.954 microseconds,
using the algorithm

	now = (timebase - tb_orig_stamp) * tb_to_xs + stamp_xsec

and then converts the time in xsecs to seconds and microseconds.

The kernel updates the tb_orig_stamp and stamp_xsec values every
tick in update_vsyscall().  If the length of the tick is not an
integer number of xsecs, then some precision is lost in converting
the current time to xsecs.  For example, with CONFIG_HZ=1000, the
tick is 1ms long, which is 1048.576 xsecs.  That means that
stamp_xsec will advance by either 1048 or 1049 on each tick.
With the right conditions, it is possible for userspace to get
(timebase - tb_orig_stamp) * tb_to_xs being 1049 if the kernel is
slightly late in updating the vdso_datapage, and then for stamp_xsec
to advance by 1048 when the kernel does update it, and for userspace
to then see (timebase - tb_orig_stamp) * tb_to_xs being zero due to
integer truncation.  The result is that time appears to go backwards
by 1 microsecond.

To fix this we change the VDSO gettimeofday to use a new field in the
VDSO datapage which stores the nanoseconds part of the time as a
fractional number of seconds in a 0.32 binary fraction format.
(Or put another way, as a 32-bit number in units of 0.23283 ns.)
This is convenient because we can use the mulhwu instruction to
convert it to either microseconds or nanoseconds.

Since it turns out that computing the time of day using this new field
is simpler than either using stamp_xsec (as gettimeofday does) or
stamp_xtime.tv_nsec (as clock_gettime does), this converts both
gettimeofday and clock_gettime to use the new field.  The existing
__do_get_tspec function is converted to use the new field and take
a parameter in r7 that indicates the desired resolution, 1,000,000
for microseconds or 1,000,000,000 for nanoseconds.  The __do_get_xsec
function is then unused and is deleted.

The new algorithm is

	now = ((timebase - tb_orig_stamp) << 12) * tb_to_xs
		+ (stamp_xtime_seconds << 32) + stamp_sec_fraction

with 'now' in units of 2^-32 seconds.  That is then converted to
seconds and either microseconds or nanoseconds with

	seconds = now >> 32
	partseconds = ((now & 0xffffffff) * resolution) >> 32

The 32-bit VDSO code also makes a further simplification: it ignores
the bottom 32 bits of the tb_to_xs value, which is a 0.64 format binary
fraction.  Doing so gets rid of 4 multiply instructions.  Assuming
a timebase frequency of 1GHz or less and an update interval of no
more than 10ms, the upper 32 bits of tb_to_xs will be at least
4503599, so the error from ignoring the low 32 bits will be at most
2.2ns, which is more than an order of magnitude less than the time
taken to do gettimeofday or clock_gettime on our fastest processors,
so there is no possibility of seeing inconsistent values due to this.

This also moves update_gtod() down next to its only caller, and makes
update_vsyscall use the time passed in via the wall_time argument rather
than accessing xtime directly.  At present, wall_time always points to
xtime, but that could change in future.

Signed-off-by: Paul Mackerras 
Signed-off-by: Benjamin Herrenschmidt