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 <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Replace EXTRA_CFLAGS with ccflags-y and EXTRA_AFLAGS with asflags-y.
Signed-off-by: matt mooney <mfm@muteddisk.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
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 <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Due to missing segment assignments the .text section was put in the NOTES
segment (and marked as NOTE section), and the .got was put in the DYNAMIC
segment.
Signed-off-by: Andreas Schwab <schwab@linux-m68k.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This patch changes the remaining direct references to
.data.page_aligned in C and assembly code to use the macros in
include/linux/linkage.h.
Signed-off-by: Tim Abbott <tabbott@ksplice.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Haavard Skinnemoen <hskinnemoen@atmel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
ld-option is misnamed as it test options to gcc, not to ld.
Renamed it to reflect this.
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Make it possible to enable GCOV code coverage measurement on powerpc.
Lightly tested on 64-bit, seems to work as expected.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We're soon running out of CPU features and I need to add some new
ones for various MMU related bits, so this patch separates the MMU
features from the CPU features. I moved over the 32-bit MMU related
ones, added base features for MMU type families, but didn't move
over any 64-bit only feature yet.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Currently the clock_gettime implementation in the VDSO produces a
result with microsecond resolution for the cases that are handled
without a system call, i.e. CLOCK_REALTIME and CLOCK_MONOTONIC. The
nanoseconds field of the result is obtained by computing a
microseconds value and multiplying by 1000.
This changes the code in the VDSO to do the computation for
clock_gettime with nanosecond resolution. That means that the
resolution of the result will ultimately depend on the timebase
frequency.
Because the timestamp in the VDSO datapage (stamp_xsec, the real time
corresponding to the timebase count in tb_orig_stamp) is in units of
2^-20 seconds, it doesn't have sufficient resolution for computing a
result with nanosecond resolution. Therefore this adds a copy of
xtime to the VDSO datapage and updates it in update_gtod() along with
the other time-related fields.
Signed-off-by: Paul Mackerras <paulus@samba.org>
To allow for a single kernel image on e500 v1/v2/mc we need to fixup lwsync
at runtime. On e500v1/v2 lwsync causes an illop so we need to patch up
the code. We default to 'sync' since that is always safe and if the cpu
is capable we will replace 'sync' with 'lwsync'.
We introduce CPU_FTR_LWSYNC as a way to determine at runtime if this is
needed. This flag could be moved elsewhere since we dont really use it
for the normal CPU_FTR purpose.
Finally we only store the relative offset in the fixup section to keep it
as small as possible rather than using a full fixup_entry.
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The current feature section logic only supports nop'ing out code, this means
if you want to choose at runtime between instruction sequences, one or both
cases will have to execute the nop'ed out contents of the other section, eg:
BEGIN_FTR_SECTION
or 1,1,1
END_FTR_SECTION_IFSET(FOO)
BEGIN_FTR_SECTION
or 2,2,2
END_FTR_SECTION_IFCLR(FOO)
and the resulting code will be either,
or 1,1,1
nop
or,
nop
or 2,2,2
For small code segments this is fine, but for larger code blocks and in
performance criticial code segments, it would be nice to avoid the nops.
This commit starts to implement logic to allow the following:
BEGIN_FTR_SECTION
or 1,1,1
FTR_SECTION_ELSE
or 2,2,2
ALT_FTR_SECTION_END_IFSET(FOO)
and the resulting code will be:
or 1,1,1
or,
or 2,2,2
We achieve this by extending the existing FTR macros. The current feature
section semantic just becomes a special case, ie. if the else case is empty
we nop out the default case.
The key limitation is that the size of the else case must be less than or
equal to the size of the default case. If the else case is smaller the
remainder of the section is nop'ed.
We let the linker put the else case code in with the rest of the text,
so that relative branches from the else case are more likley to link,
this has the disadvantage that we can't free the unused else cases.
This commit introduces the required macro and linker script changes, but
does not enable the patching of the alternative sections.
We also need to update two hand-made section entries in reg.h and timex.h
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The current VDSO implementation is hardcoded to 128 byte cache blocks,
which are only used on IBM's 64-bit processors.
Convert it to get the cache block sizes out of vdso_data instead,
similar to how the ppc64 in-kernel cache flush does it.
Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This cleans up the formatting in the vDSO linker script, mostly just the
use of whitespace. It's intended to approximate the kernel standard
conventions for indenting C, treating elements of the linker script about
like initialized variable definitions.
Signed-off-by: Roland McGrath <roland@redhat.com>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This keeps an unstripped copy of the vDSO images built before they are
stripped and embedded in the kernel. The unstripped copies get installed in
$(MODLIB)/vdso/ by "make install". These files can be useful when they
contain source-level debugging information.
Signed-off-by: Roland McGrath <roland@redhat.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Consider the prototype for gettimeofday():
int gettimofday(struct timeval *tv, struct timezone *tz);
Although it is valid to call with /either/ tv or tz being NULL, and
the C version of sys_gettimeofday() supports this, the current version
of gettimeofday() in the VDSO will SEGV if called with a NULL tv.
This adds a check for tv being NULL so that it doesn't SEGV.
Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This patch reworks the feature fixup mecanism so vdso's can be fixed up.
The main issue was that the construct:
.long label (or .llong on 64 bits)
will not work in the case of a shared library like the vdso. It will
generate an empty placeholder in the fixup table along with a reloc,
which is not something we can deal with in the vdso.
The idea here (thanks Alan Modra !) is to instead use something like:
1:
.long label - 1b
That is, the feature fixup tables no longer contain addresses of bits of
code to patch, but offsets of such code from the fixup table entry
itself. That is properly resolved by ld when building the .so's. I've
modified the fixup mecanism generically to use that method for the rest
of the kernel as well.
Another trick is that the 32 bits vDSO included in the 64 bits kernel
need to have a table in the 64 bits format. However, gas does not
support 32 bits code with a statement of the form:
.llong label - 1b (Or even just .llong label)
That is, it cannot emit the right fixup/relocation for the linker to use
to assign a 32 bits address to an .llong field. Thus, in the specific
case of the 32 bits vdso built as part of the 64 bits kernel, we are
using a modified macro that generates:
.long 0xffffffff
.llong label - 1b
Note that is assumes that the value is negative which is enforced by
the .lds (those offsets are always negative as the .text is always
before the fixup table and gas doesn't support emiting the reloc the
other way around).
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The latest toolchains can produce a new ELF section in DSOs and
dynamically-linked executables. The new section ".gnu.hash" replaces
".hash", and allows for more efficient runtime symbol lookups by the
dynamic linker. The new ld option --hash-style={sysv|gnu|both} controls
whether to produce the old ".hash", the new ".gnu.hash", or both. In some
new systems such as Fedora Core 6, gcc by default passes --hash-style=gnu
to the linker, so that a standard invocation of "gcc -shared" results in
producing a DSO with only ".gnu.hash". The new ".gnu.hash" sections need
to be dealt with the same way as ".hash" sections in all respects; only the
dynamic linker cares about their contents. To work with older dynamic
linkers (i.e. preexisting releases of glibc), a binary must have the old
".hash" section. The --hash-style=both option produces binaries that a new
dynamic linker can use more efficiently, but an old dynamic linker can
still handle.
The new section runs afoul of the custom linker scripts used to build vDSO
images for the kernel. On ia64, the failure mode for this is a boot-time
panic because the vDSO's PT_IA_64_UNWIND segment winds up ill-formed.
This patch addresses the problem in two ways.
First, it mentions ".gnu.hash" in all the linker scripts alongside ".hash".
This produces correct vDSO images with --hash-style=sysv (or old tools),
with --hash-style=gnu, or with --hash-style=both.
Second, it passes the --hash-style=sysv option when building the vDSO
images, so that ".gnu.hash" is not actually produced. This is the most
conservative choice for compatibility with any old userland. There is some
concern that some ancient glibc builds (though not any known old production
system) might choke on --hash-style=both binaries. The optimizations
provided by the new style of hash section do not really matter for a DSO
with a tiny number of symbols, as the vDSO has. If someone wants to use
=gnu or =both for their vDSO builds and worry less about that
compatibility, just change the option and the linker script changes will
make any choice work fine.
Signed-off-by: Roland McGrath <roland@redhat.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Andi Kleen <ak@muc.de>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Mark unwind info for signal trampolines using the new S augmentation flag
introduced in: http://gcc.gnu.org/PR26208.
GCC 4.2 (or patched earlier GCC) will be able to special case unwinding
through frames right above signal trampolines. As the augmentations start
with z flag and S is at the very end of the augmentation string, older GCCs
will just skip the S flag as unknown (that's why an augmentation flag was
chosen over say a new CFA opcode).
Signed-off-by: Jakub Jelinek <jakub@redhat.com>
Cc: Andi Kleen <ak@muc.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The vDSO functions should have the same calling convention as a syscall.
Unfortunately, they currently don't set the cr0.so bit which is used to
indicate an error. This patch makes them clear this bit unconditionally
since all functions currently succeed. The syscall fallback done by some
of them will eventually override this if the syscall fails.
This also changes the symbol version of all vdso exports to make sure
glibc can differenciate between old and fixed calls for existing ones
like __kernel_gettimeofday.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This fixes various errors in the new functions added in the vDSO's,
I've now verified all functions on both 32 and 64 bits vDSOs. It also
fix a sign extension bug getting the initial time of day at boot that
could cause the monotonic clock value to be completely on bogus for
64 bits applications (with either the vDSO or the syscall) on
powermacs.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
We needed the VDSO symbols in the arch/ppc asm-offsets.c, and there
were a few usages of _systemcfg still left lying around.
Signed-off-by: Paul Mackerras <paulus@samba.org>
This patch moves the vdso's to arch/powerpc, adds support for the 32
bits vdso to the 32 bits kernel, rename systemcfg (finally !), and adds
some new (still untested) routines to both vdso's: clock_gettime() with
support for CLOCK_REALTIME and CLOCK_MONOTONIC, clock_getres() (same
clocks) and get_tbfreq() for glibc to retreive the timebase frequency.
Tom,Steve: The implementation of get_tbfreq() I've done for 32 bits
returns a long long (r3, r4) not a long. This is such that if we ever
add support for >4Ghz timebases on ppc32, the userland interface won't
have to change.
I have tested gettimeofday() using some glibc patches in both ppc32 and
ppc64 kernels using 32 bits userland (I haven't had a chance to test a
64 bits userland yet, but the implementation didn't change and was
tested earlier). I haven't tested yet the new functions.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Anton Blanchard