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: NPaul Mackerras <paulus@samba.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Now we dynamically allocate the paca array, it takes an extra load
whenever we want to access another cpu's paca. One place we do that a lot
is per cpu variables. A simple example:

DEFINE_PER_CPU(unsigned long, vara);
unsigned long test4(int cpu)
{
	return per_cpu(vara, cpu);
}

This takes 4 loads, 5 if you include the actual load of the per cpu variable:

    ld r11,-32760(r30)  # load address of paca pointer
    ld r9,-32768(r30)   # load link address of percpu variable
    sldi r3,r29,9       # get offset into paca (each entry is 512 bytes)
    ld r0,0(r11)        # load paca pointer
    add r3,r0,r3        # paca + offset
    ld r11,64(r3)       # load paca[cpu].data_offset

    ldx r3,r9,r11       # load per cpu variable

If we remove the ppc64 specific per_cpu_offset(), we get the generic one
which indexes into a statically allocated array. This removes one load and
one add:

    ld r11,-32760(r30)  # load address of __per_cpu_offset
    ld r9,-32768(r30)   # load link address of percpu variable
    sldi r3,r29,3       # get offset into __per_cpu_offset (each entry 8 bytes)
    ldx r11,r11,r3      # load __per_cpu_offset[cpu]

    ldx r3,r9,r11       # load per cpu variable

Having all the offsets in one array also helps when iterating over a per cpu
variable across a number of cpus, such as in the scheduler. Before we would
need to load one paca cacheline when calculating each per cpu offset. Now we
have 16 (128 / sizeof(long)) per cpu offsets in each cacheline.
Signed-off-by: NAnton Blanchard <anton@samba.org>
Signed-off-by: NBenjamin 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: NPaul Mackerras <paulus@samba.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

In kexec_prepare_cpus, the primary CPU IPIs the secondary CPUs to
kexec_smp_down().  kexec_smp_down() calls kexec_smp_wait() which sets
the hw_cpu_id() to -1.  The primary does this while leaving IRQs on
which means the primary can take a timer interrupt which can lead to
the IPIing one of the secondary CPUs (say, for a scheduler re-balance)
but since the secondary CPU now has a hw_cpu_id = -1, we IPI CPU
-1... Kaboom!

We are hitting this case regularly on POWER7 machines.

There is also a second race, where the primary will tear down the MMU
mappings before knowing the secondaries have entered real mode.

Also, the secondaries are clearing out any pending IPIs before
guaranteeing that no more will be received.

This changes kexec_prepare_cpus() so that we turn off IRQs in the
primary CPU much earlier.  It adds a paca flag to say that the
secondaries have entered the kexec_smp_down() IPI and turned off IRQs,
rather than overloading hw_cpu_id with -1.  This new paca flag is
again used to in indicate when the secondaries has entered real mode.

It also ensures that all CPUs have their IRQs off before we clear out
any pending IPI requests (in kexec_cpu_down()) to ensure there are no
trailing IPIs left unacknowledged.
Signed-off-by: NMichael Neuling <mikey@neuling.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

When we build with ftrace enabled its possible that loadcam_entry would
have used the stack pointer (even though the code doesn't need it).  We
call loadcam_entry in __secondary_start before the stack is setup.  To
ensure that loadcam_entry doesn't use the stack pointer the easiest
solution is to just have it in asm code.
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

We need the SWITCH_FRAME_SIZE define on Book3S_32 now too.
So let's export it unconditionally.

CC: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NAlexander Graf <agraf@suse.de>
Acked-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NAvi Kivity <avi@redhat.com>

We need to keep the pointer to the shadow vcpu somewhere accessible from
within really early interrupt code. The best fit I found was the thread
struct, as that resides in an SPRG.

So let's put a pointer to the shadow vcpu in the thread struct and add
an asm-offset so we can find it.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

The shadow vcpu now contains some fields we don't use from the vcpu anymore.
Access to them happens using inline functions that happily use the shadow
vcpu fields.

So let's now ifdef them out to booke only and add asm-offsets.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

Upstream recently added a new name for PPC64: Book3S_64.

So instead of using CONFIG_PPC64 we should use CONFIG_PPC_BOOK3S consotently.
That makes understanding the code easier (I hope).
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

When we build with ftrace enabled its possible that loadcam_entry would
have used the stack pointer (even though the code doesn't need it).  We
call loadcam_entry in __secondary_start before the stack is setup.  To
ensure that loadcam_entry doesn't use the stack pointer the easiest
solution is to just have it in asm code.
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

Anton Blanchard found that large POWER systems would occasionally
crash in the exception exit path when profiling with perf_events.
The symptom was that an interrupt would occur late in the exit path
when the MSR[RI] (recoverable interrupt) bit was clear.  Interrupts
should be hard-disabled at this point but they were enabled.  Because
the interrupt was not recoverable the system panicked.

The reason is that the exception exit path was calling
perf_event_do_pending after hard-disabling interrupts, and
perf_event_do_pending will re-enable interrupts.

The simplest and cleanest fix for this is to use the same mechanism
that 32-bit powerpc does, namely to cause a self-IPI by setting the
decrementer to 1.  This means we can remove the tests in the exception
exit path and raw_local_irq_restore.

This also makes sure that the call to perf_event_do_pending from
timer_interrupt() happens within irq_enter/irq_exit.  (Note that
calling perf_event_do_pending from timer_interrupt does not mean that
there is a possible 1/HZ latency; setting the decrementer to 1 ensures
that the timer interrupt will happen immediately, i.e. within one
timebase tick, which is a few nanoseconds or 10s of nanoseconds.)
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Cc: stable@kernel.org
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

SRR1 stores more information that just the MSR value. It also stores
valuable information about the type of interrupt we received, for
example whether the storage interrupt we just got was because of a
missing htab entry or not.

We use that information to speed up the exit path.

Now if we get preempted before we can interpret the shadow_msr values,
we get into vcpu_put which then calls the MSR handler, which then sets
all the SRR1 information bits in shadow_msr to 0. Great.

So let's preserve the SRR1 specific bits in shadow_msr whenever we set
the MSR. They don't hurt.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

Currently we're racy when doing the transition from IR=1 to IR=0, from
the module memory entry code to the real mode SLB switching code.

To work around that I took a look at the RTAS entry code which is faced
with a similar problem and did the same thing:

  A small helper in linear mapped memory that does mtmsr with IR=0 and
  then RFIs info the actual handler.

Thanks to that trick we can safely take page faults in the entry code
and only need to be really wary of what to do as of the SLB switching
part.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

We're being horribly racy right now. All the entry and exit code hijacks
random fields from the PACA that could easily be used by different code in
case we get interrupted, for example by a #MC or even page fault.

After discussing this with Ben, we figured it's best to reserve some more
space in the PACA and just shove off some vcpu state to there.

That way we can drastically improve the readability of the code, make it
less racy and less complex.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAvi Kivity <avi@redhat.com>

Re-write the code so its more standalone and fixed some issues:
* Bump'd # of CAM entries to 64 to support e500mc
* Make the code handle MAS7 properly
* Use pr_cont instead of creating a string as we go
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

In order to access fields in the PACA from assembly code, we need
to generate offsets using asm-offsets.c.

So let's add the new PACA related bits, we just introduced!
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

We need to access some VCPU fields from assembly code. In order to get
the proper offsets, we have to define them in asm-offsets.c.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Bye-bye Performance Counters, welcome Performance Events!

In the past few months the perfcounters subsystem has grown out its
initial role of counting hardware events, and has become (and is
becoming) a much broader generic event enumeration, reporting, logging,
monitoring, analysis facility.

Naming its core object 'perf_counter' and naming the subsystem
'perfcounters' has become more and more of a misnomer. With pending
code like hw-breakpoints support the 'counter' name is less and
less appropriate.

All in one, we've decided to rename the subsystem to 'performance
events' and to propagate this rename through all fields, variables
and API names. (in an ABI compatible fashion)

The word 'event' is also a bit shorter than 'counter' - which makes
it slightly more convenient to write/handle as well.

Thanks goes to Stephane Eranian who first observed this misnomer and
suggested a rename.

User-space tooling and ABI compatibility is not affected - this patch
should be function-invariant. (Also, defconfigs were not touched to
keep the size down.)

This patch has been generated via the following script:

  FILES=$(find * -type f | grep -vE 'oprofile|[^K]config')

  sed -i \
    -e 's/PERF_EVENT_/PERF_RECORD_/g' \
    -e 's/PERF_COUNTER/PERF_EVENT/g' \
    -e 's/perf_counter/perf_event/g' \
    -e 's/nb_counters/nb_events/g' \
    -e 's/swcounter/swevent/g' \
    -e 's/tpcounter_event/tp_event/g' \
    $FILES

  for N in $(find . -name perf_counter.[ch]); do
    M=$(echo $N | sed 's/perf_counter/perf_event/g')
    mv $N $M
  done

  FILES=$(find . -name perf_event.*)

  sed -i \
    -e 's/COUNTER_MASK/REG_MASK/g' \
    -e 's/COUNTER/EVENT/g' \
    -e 's/\<event\>/event_id/g' \
    -e 's/counter/event/g' \
    -e 's/Counter/Event/g' \
    $FILES

... to keep it as correct as possible. This script can also be
used by anyone who has pending perfcounters patches - it converts
a Linux kernel tree over to the new naming. We tried to time this
change to the point in time where the amount of pending patches
is the smallest: the end of the merge window.

Namespace clashes were fixed up in a preparatory patch - and some
stylistic fallout will be fixed up in a subsequent patch.

( NOTE: 'counters' are still the proper terminology when we deal
  with hardware registers - and these sed scripts are a bit
  over-eager in renaming them. I've undone some of that, but
  in case there's something left where 'counter' would be
  better than 'event' we can undo that on an individual basis
  instead of touching an otherwise nicely automated patch. )
Suggested-by: NStephane Eranian <eranian@google.com>
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: NPaul Mackerras <paulus@samba.org>
Reviewed-by: NArjan van de Ven <arjan@linux.intel.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: <linux-arch@vger.kernel.org>
LKML-Reference: <new-submission>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

This adds various fields in the PACA that are for use specifically
by Book3E processors, such as exception save areas, current pgd
pointer, special exceptions kernel stacks etc...
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

This adds the PTE and pgtable format definitions, along with changes
to the kernel memory map and other definitions related to implementing
support for 64-bit Book3E. This also shields some asm-offset bits that
are currently only relevant on 32-bit

We also move the definition of the "linux" page size constants to
the common mmu.h file and add a few sizes that are relevant to
embedded processors.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

This provides a mechanism to allow the perf_counters code to access
user memory in a PMU interrupt routine.  Such an access can cause
various kinds of interrupt: SLB miss, MMU hash table miss, segment
table miss, or TLB miss, depending on the processor.  This commit
only deals with 64-bit classic/server processors, which use an MMU
hash table.  32-bit processors are already able to access user memory
at interrupt time.  Since we don't soft-disable on 32-bit, we avoid
the possibility of reentering hash_page or the TLB miss handlers,
since they run with interrupts disabled.

On 64-bit processors, an SLB miss interrupt on a user address will
update the slb_cache and slb_cache_ptr fields in the paca.  This is
OK except in the case where a PMU interrupt occurs in switch_slb,
which also accesses those fields.  To prevent this, we hard-disable
interrupts in switch_slb.  Interrupts are already soft-disabled at
this point, and will get hard-enabled when they get soft-enabled
later.

This also reworks slb_flush_and_rebolt: to avoid hard-disabling twice,
and to make sure that it clears the slb_cache_ptr when called from
other callers than switch_slb, the existing routine is renamed to
__slb_flush_and_rebolt, which is called by switch_slb and the new
version of slb_flush_and_rebolt.

Similarly, switch_stab (used on POWER3 and RS64 processors) gets a
hard_irq_disable() to protect the per-cpu variables used there and
in ste_allocate.

If a MMU hashtable miss interrupt occurs, normally we would call
hash_page to look up the Linux PTE for the address and create a HPTE.
However, hash_page is fairly complex and takes some locks, so to
avoid the possibility of deadlock, we check the preemption count
to see if we are in a (pseudo-)NMI handler, and if so, we don't call
hash_page but instead treat it like a bad access that will get
reported up through the exception table mechanism.  An interrupt
whose handler runs even though the interrupt occurred when
soft-disabled (such as the PMU interrupt) is considered a pseudo-NMI
handler, which should use nmi_enter()/nmi_exit() rather than
irq_enter()/irq_exit().
Acked-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NPaul Mackerras <paulus@samba.org>

This patch has no effect other than re-ordering PACA fields on
current server CPUs. It however is a pre-requisite for future
support of BookE 64-bit processors. Various parts of the PACA
struct are now moved under some ifdef's, either the new
CONFIG_PPC_BOOK3S or CONFIG_PPC_STD_MMU_64, whatever seems more
appropriate.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.craashing.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Signed-off-by: NLiu Yu <yu.liu@freescale.com>
Signed-off-by: NHollis Blanchard <hollisb@us.ibm.com>
Signed-off-by: NAvi Kivity <avi@redhat.com>

Signed-off-by: NMichael Ellerman <michael@ellerman.id.au>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Because 64-bit powerpc uses lazy (soft) interrupt disabling, it is
possible for a performance monitor exception to come in when the
kernel thinks interrupts are disabled (i.e. when they are
soft-disabled but hard-enabled).  In such a situation the performance
monitor exception handler might have some processing to do (such as
process wakeups) which can't be done in what is effectively an NMI
handler.

This provides a way to defer that work until interrupts get enabled,
either in raw_local_irq_restore() or by returning from an interrupt
handler to code that had interrupts enabled.  We have a per-processor
flag that indicates that there is work pending to do when interrupts
subsequently get re-enabled.  This flag is checked in the interrupt
return path and in raw_local_irq_restore(), and if it is set,
perf_counter_do_pending() is called to do the pending work.
Signed-off-by: NPaul Mackerras <paulus@samba.org>

Some assembly code in head_fsl_booke.S hard-coded the size of struct tlbcam
to 20 when it indexed the TLBCAM table.  Anyone changing the size of struct
tlbcam would not know to expect that.

The kernel already has a system to get the size of C structures into
assembly language files, asm-offsets, so let's use it.

The definition of the struct gets moved to a header, so that asm-offsets.c
can include it.
Signed-off-by: NTrent Piepho <tpiepho@freescale.com>
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

Existing KVM statistics are either just counters (kvm_stat) reported for
KVM generally or trace based aproaches like kvm_trace.
For KVM on powerpc we had the need to track the timings of the different exit
types. While this could be achieved parsing data created with a kvm_trace
extension this adds too much overhead (at least on embedded PowerPC) slowing
down the workloads we wanted to measure.

Therefore this patch adds a in-kernel exit timing statistic to the powerpc kvm
code. These statistic is available per vm&vcpu under the kvm debugfs directory.
As this statistic is low, but still some overhead it can be enabled via a
.config entry and should be off by default.

Since this patch touched all powerpc kvm_stat code anyway this code is now
merged and simplified together with the exit timing statistic code (still
working with exit timing disabled in .config).
Signed-off-by: NChristian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
Signed-off-by: NHollis Blanchard <hollisb@us.ibm.com>
Signed-off-by: NAvi Kivity <avi@redhat.com>

Formerly, we used to maintain a per-vcpu shadow TLB and on every entry to the
guest would load this array into the hardware TLB. This consumed 1280 bytes of
memory (64 entries of 16 bytes plus a struct page pointer each), and also
required some assembly to loop over the array on every entry.

Instead of saving a copy in memory, we can just store shadow mappings directly
into the hardware TLB, accepting that the host kernel will clobber these as
part of the normal 440 TLB round robin. When we do that we need less than half
the memory, and we have decreased the exit handling time for all guest exits,
at the cost of increased number of TLB misses because the host overwrites some
guest entries.

These savings will be increased on processors with larger TLBs or which
implement intelligent flush instructions like tlbivax (which will avoid the
need to walk arrays in software).

In addition to that and to the code simplification, we have a greater chance of
leaving other host userspace mappings in the TLB, instead of forcing all
subsequent tasks to re-fault all their mappings.
Signed-off-by: NHollis Blanchard <hollisb@us.ibm.com>
Signed-off-by: NAvi Kivity <avi@redhat.com>

This patch doesn't yet move all 44x-specific data into the new structure, but
is the first step down that path. In the future we may also want to create a
struct kvm_vcpu_booke.

Based on patch from Liu Yu <yu.liu@freescale.com>.
Signed-off-by: NHollis Blanchard <hollisb@us.ibm.com>
Signed-off-by: NAvi Kivity <avi@redhat.com>

This will ease ports to other cores.

Also remove unused "struct kvm_tlb" while we're at it.
Signed-off-by: NHollis Blanchard <hollisb@us.ibm.com>
Signed-off-by: NAvi Kivity <avi@redhat.com>

This adds support for 16k and 64k page sizes on PowerPC 44x processors.

The PGDIR table is much smaller than a page when using 16k or 64k
pages (512 and 32 bytes respectively) so we allocate the PGDIR with
kzalloc() instead of __get_free_pages().

One PTE table covers rather a large memory area when using 16k or 64k
pages (32MB or 512MB respectively), so we can easily put FIXMAP and
PKMAP in the area covered by one PTE table.
Signed-off-by: NYuri Tikhonov <yur@emcraft.com>
Signed-off-by: NVladimir Panfilov <pvr@emcraft.com>
Signed-off-by: NIlya Yanok <yanok@emcraft.com>
Acked-by: NJosh Boyer <jwboyer@linux.vnet.ibm.com>
Signed-off-by: NPaul Mackerras <paulus@samba.org>

This splits the mmu_context handling between 32-bit hash based
processors, 64-bit hash based processors and everybody else.  This is
preliminary work for adding SMP support for BookE processors.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: NKumar Gala <galak@kernel.crashing.org>
Signed-off-by: NPaul 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: NPaul Mackerras <paulus@samba.org>

When we use TID=N userspace mappings, we must ensure that kernel mappings have
been destroyed when entering userspace. Using TID=1/TID=0 for kernel/user
mappings and running userspace with PID=0 means that userspace can't access the
kernel mappings, but the kernel can directly access userspace.

The net is that we don't need to flush the TLB on privilege switches, but we do
on guest context switches (which are far more infrequent). Guest boot time
performance improvement: about 30%.
Signed-off-by: NHollis Blanchard <hollisb@us.ibm.com>
Signed-off-by: NAvi Kivity <avi@qumranet.com>

Track which TLB entries need to be written, instead of overwriting everything
below the high water mark. Typically only a single guest TLB entry will be
modified in a single exit.

Guest boot time performance improvement: about 15%.
Signed-off-by: NHollis Blanchard <hollisb@us.ibm.com>
Signed-off-by: NAvi Kivity <avi@qumranet.com>

We're saving the host TLB state to memory on every exit, but never using it.
Originally I had thought that we'd want to restore host TLB for heavyweight
exits, but that could actually hurt when context switching to an unrelated host
process (i.e. not qemu).

Since this decreases the performance penalty of all exits, this patch improves
guest boot time by about 15%.
Signed-off-by: NHollis Blanchard <hollisb@us.ibm.com>
Signed-off-by: NAvi Kivity <avi@qumranet.com>

This rearranges a bit of code, and adds support for
36-bit physical addressing for configs that use a
hashed page table.  The 36b physical support is not
enabled by default on any config - it must be
explicitly enabled via the config system.

This patch *only* expands the page table code to accomodate
large physical addresses on 32-bit systems and enables the
PHYS_64BIT config option for 86xx.  It does *not*
allow you to boot a board with more than about 3.5GB of
RAM - for that, SWIOTLB support is also required (and
coming soon).
Signed-off-by: NBecky Bruce <becky.bruce@freescale.com>
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

This changes the way that the exception prologs transfer control to
the handlers in 64-bit kernels with the aim of making it possible to
have the prologs separate from the main body of the kernel.  Now,
instead of computing the address of the handler by taking the top
32 bits of the paca address (to get the 0xc0000000........ part) and
ORing in something in the bottom 16 bits, we get the base address of
the kernel by doing a load from the paca and add an offset.

This also replaces an mfmsr and an ori to compute the MSR value for
the handler with a load from the paca.  That makes it unnecessary to
have a separate version of EXCEPTION_PROLOG_PSERIES that forces 64-bit
mode.

We can no longer use a direct branches in the exception prolog code,
which means that the SLB miss handlers can't branch directly to
.slb_miss_realmode any more.  Instead we have to compute the address
and do an indirect branch.  This is conditional on CONFIG_RELOCATABLE;
for non-relocatable kernels we use a direct branch as before.  (A later
change will allow CONFIG_RELOCATABLE to be set on 64-bit powerpc.)

Since the secondary CPUs on pSeries start execution in the first 0x100
bytes of real memory and then have to get to wherever the kernel is,
we can't use a direct branch to get there.  Instead this changes
__secondary_hold_spinloop from a flag to a function pointer.  When it
is set to a non-NULL value, the secondary CPUs jump to the function
pointed to by that value.

Finally this eliminates one code difference between 32-bit and 64-bit
by making __secondary_hold be the text address of the secondary CPU
spinloop rather than a function descriptor for it.
Signed-off-by: NPaul Mackerras <paulus@samba.org>

The layout of the new VSR registers and how they overlap on top of the
legacy FPR and VR registers is:

                   VSR doubleword 0               VSR doubleword 1
          ----------------------------------------------------------------
  VSR[0]  |             FPR[0]            |                              |
          ----------------------------------------------------------------
  VSR[1]  |             FPR[1]            |                              |
          ----------------------------------------------------------------
          |              ...              |                              |
          |              ...              |                              |
          ----------------------------------------------------------------
  VSR[30] |             FPR[30]           |                              |
          ----------------------------------------------------------------
  VSR[31] |             FPR[31]           |                              |
          ----------------------------------------------------------------
  VSR[32] |                             VR[0]                            |
          ----------------------------------------------------------------
  VSR[33] |                             VR[1]                            |
          ----------------------------------------------------------------
          |                              ...                             |
          |                              ...                             |
          ----------------------------------------------------------------
  VSR[62] |                             VR[30]                           |
          ----------------------------------------------------------------
  VSR[63] |                             VR[31]                           |
          ----------------------------------------------------------------

VSX has 64 128bit registers.  The first 32 regs overlap with the FP
registers and hence extend them with and additional 64 bits.  The
second 32 regs overlap with the VMX registers.

This commit introduces the thread_struct changes required to reflect
this register layout.  Ptrace and signals code is updated so that the
floating point registers are correctly accessed from the thread_struct
when CONFIG_VSX is enabled.
Signed-off-by: NMichael Neuling <mikey@neuling.org>
Signed-off-by: NPaul Mackerras <paulus@samba.org>

On machines with more than one exception level any system register that
might be modified by the "normal" exception level needs to be saved and
restored on taking a higher level exception.  We already are saving
and restoring ESR and DEAR.

For critical level add SRR0/1.
For debug level add CSRR0/1 and SRR0/1.
For machine check level add DSRR0/1, CSRR0/1, and SRR0/1.

On FSL Book-E parts we always save/restore the MAS registers for critical,
debug, and machine check level exceptions.  On 44x we always save/restore
the MMUCR.

Additionally, we save and restore the ksp_limit since we have to adjust it
for each exception level.
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>
Acked-by: NPaul Mackerras <paulus@samba.org>