There is a tricky interaction between the machine check handler
and the critical sections of load_fpu_regs and save_fpu_regs
functions. If the machine check interrupts one of the two
functions the critical section cleanup will complete the function
before the machine check handler s390_do_machine_check is called.
Trouble is that the machine check handler needs to validate the
floating point registers *before* and not *after* the completion
of load_fpu_regs/save_fpu_regs.

The simplest solution is to rewind the PSW to the start of the
load_fpu_regs/save_fpu_regs and retry the function after the
return from the machine check handler.
Tested-by: NChristian Borntraeger <borntraeger@de.ibm.com>
Cc: <stable@vger.kernel.org> # 4.3+
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Add the missing lpp magic initialization for cpu 0. Without this all
samples on cpu 0 do not have the most significant bit set in the
program parameter field, which we use to distinguish between guest and
host samples if the pid is also 0.

We did initialize the lpp magic in the absolute zero lowcore but
forgot that when switching to the allocated lowcore on cpu 0 only.
Reported-by: NShu Juan Zhang <zhshuj@cn.ibm.com>
Acked-by: NChristian Borntraeger <borntraeger@de.ibm.com>
Cc: stable@vger.kernel.org # v4.4+
Fixes: e22cf8ca ("s390/cpumf: rework program parameter setting to detect guest samples")
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Signed-off-by: NAdam Buchbinder <adam.buchbinder@gmail.com>
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The pgtable.c file is quite big, before it grows any larger split it
into pgtable.c, pgalloc.c and gmap.c. In addition move the gmap related
header definitions into the new gmap.h header and all of the pgste
helpers from pgtable.h to pgtable.c.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The pmdp_xxx function are smaller than their ptep_xxx counterparts
but to keep things symmetrical unline them as well.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The code in the various ptep_xxx functions has grown quite large,
consolidate them to four out-of-line functions:
  ptep_xchg_direct to exchange a pte with another with immediate flushing
  ptep_xchg_lazy to exchange a pte with another in a batched update
  ptep_modify_prot_start to begin a protection flags update
  ptep_modify_prot_commit to commit a protection flags update
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Provide a user space interface to issue call logical-processor instructions.
Only selected CLP commands are allowed, enough to get the full overview of
the installed PCI functions.
Reviewed-by: NSebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Convert the uses of pr_warning to pr_warn so there are fewer
uses of the old pr_warning.

Miscellanea:

o Align arguments
o Coalesce formats
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

git commit 26f15caa ("s390/cmpxchg: simplify cmpxchg_double")
removed support for cmpxchg_double for two consecutive four byte
values, for which it would generate a cds instruction.

However I forgot to remove the corresponding define in our percpu
header file, which means that this_cpu_cmpxchg_double would now
incorrectly generate a cdsg instruction if being used on a double four
byte location. Therefore remove the percpu define as well.

There is currently no user and therefore no bug fixed with
this. Obviously any such user could and should simply use cmpxchg.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

commit e22cf8ca ("s390/cpumf: rework program parameter setting
to detect guest samples") requires guest changes to get proper
guest/host. We can do better: We can use the primary asn value,
which is set on all Linux variants to compare this with the host
pp value.
We now have the following cases:
1. Guest using PP
host sample:  gpp == 0, asn == hpp --> host
guest sample: gpp != 0 --> guest
2. Guest not using PP
host sample:  gpp == 0, asn == hpp --> host
guest sample: gpp == 0, asn != hpp --> guest

As soon as the host no longer sets CR4, we must back out
this heuristics - let's add a comment in switch_to.
Signed-off-by: NChristian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: NHendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

We have two close to identical report_user_fault functions.
Add a parameter to one and get rid of the other one in order
to reduce code duplication.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The double escape character sequence introduced with commit
272fa59c ("s390/dis: Fix handling of format specifiers") is not
necessary anymore since commit 561e1030 ("s390/dis: Fix printing
of the register numbers").

Instead this now generates an extra '%' character:

lg      %%r1,160(%%r11)

So fix this and basically revert 272fa59c.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Git commit ab3f285f
"KVM: s390/mm: try a cow on read only pages for key ops"
added a fixup_user_fault to set_guest_storage_key force a copy on
write if the page is mapped read-only. This is supposed to fix the
problem of differing storage keys for shared mappings, e.g. the
empty_zero_page.
But if the storage key is set before the pte is mapped the storage
key update is done on the pgste. A later fault will happily map the
shared page with the key from the pgste.

Eventually git commit 2faee8ff
"s390/mm: prevent and break zero page mappings in case of storage keys"
fixed this problem for the empty_zero_page. The commit makes sure that
guests enabled for storage keys will not use the empty_zero_page at all.

As the call to fixup_user_fault in set_guest_storage_key depends on the
order of the storage key operation vs. the fault that maps the pte
it does not really fix anything. Just remove it.
Reviewed-by: NDominik Dingel <dingel@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The comment describing the bit encoding for segment table entries
is incorrect in regard to the read and write bits. The segment
read bit is 0x0002 and write is 0x0001, not the other way around.
Reported-by: NGerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

We have four different stack tracers of which three had bugs. So it's
time to merge them to a single stack tracer which allows to specify a
call back function which will be called for each step.

This patch changes behavior a bit:

- the "nosched" and "in_sched_functions" check within
  save_stack_trace_tsk did work only for the last stack frame within a
  context. Now it considers the check for each stack frame like it
  should.

- both the oprofile variant and the perf_events variant did save a
  return address twice if a zero back chain was detected, which
  indicates an interrupt frame. The new dump_trace function will call
  the oprofile and perf_events backends with the psw address that is
  contained within the corresponding pt_regs structure instead.

- the original show_trace and save_context_stack functions did already
  use the psw address of the pt_regs structure if a zero back chain
  was detected. However now we ignore the psw address if it is a user
  space address. After all we trace the kernel stack and not the user
  space stack. This way we also get rid of the garbage user space
  address in case of warnings and / or panic call traces.

So this should make life easier since now there is only one stack
tracer left which we can break.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The first parameter of pgste_update_all is a pointer to a pte.
Simplify the code by passing the pte value.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Implement current_stack_pointer() helper function and use it
everywhere, instead of having several different inline assembly
variants.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Tested-by: NPeter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Use the inversed "nosched" logic like all other architectures.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Tested-by: NPeter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The change of the access rights for an address range in the kernel
address space is currently done with a loop of IPTE + a store of the
modified PTE. Between the IPTE and the store the PTE will be invalid,
this intermediate state can cause problems with concurrent accesses.

Consider a change of a kernel area from read-write to read-only, a
concurrent reader of that area should be fine but with the invalid
PTE it might get an unexpected exception.

Remove the IPTEs for each PTE and do a global flush after all PTEs
have been modified.
Reviewed-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Acked-by: NChristian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

For each PCI function we need to maintain arch specific data in
struct zpci_dev which also contains a pointer to struct pci_dev.

When a function is registered or deregistered (which is triggered by PCI
common code) we need to adjust that pointer which could interfere with
the machine check handler (triggered by FW) using zpci_dev->pdev.

Since multiple instances of the same pdev could exist at a time this can't
be solved with locking.

Fix that by ditching the pdev pointer and use a bus walk to reach
struct pci_dev (only one instance of a pdev can be registered at the bus
at a time).
Signed-off-by: NSebastian Ott <sebott@linux.vnet.ibm.com>
Reviewed-by: NGerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

git commit 904818e2
"s390/kernel: introduce fpu-internal.h with fpu helper functions"
introduced the fpregs_store / fp_regs_load helper. These function
fail to save and restore the floating pointer control registers.

The effect is that the FPC is not correctly handled on signal
delivery and signal return.

Cc: stable@vger.kernel.org # 4.4
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

git commit 80703617
"s390: add support for vector extension"
broke 31-bit compat processes in regard to signal handling.

The restore_sigregs_ext32() function is used to restore the additional
elements from the user space signal frame. Among the additional elements
are the upper registers halves for 64-bit register support for 31-bit
processes. The copy_from_user that is used to retrieve the high-gprs
array from the user stack uses an incorrect length, 8 bytes instead of
64 bytes. This causes incorrect upper register halves to get loaded.

Cc: stable@vger.kernel.org # 3.8+
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

When fixing the DAT off bug ("s390: fix DAT off memory access, e.g.
on kdump") both Christian and I missed that we can save an additional
stnsm instruction.

This saves us a couple of cycles which could improve the speed of
memcpy_real.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The diagnose tracer will indirectly call back into the lockdep code
when lockdep does not expect it (arch_spinlock). This causes lockdep
to disable itself and therefore we don't have a working lock
dependency validator anymore.

This patch effectively disables tracing of diag 0x9c and 0x44 if
lockdep is enabled.  If however lockdep is enabled spinlocks are
mainly implemented using a trylock variant, which will not issue any
diag 0x9c or 0x44. So this change has hardly any effect on tracing
except when arch_spinlock and friends are explicitly used.
Reported-and-Tested-by: NDavid Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Reviewed-by: NDavid Hildenbrand <dahi@linux.vnet.ibm.com>
Acked-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

commit 204ee2c5 ("s390/irqflags: optimize irq restore") optimized
irqrestore to really only care about interrupts and adapted the
remaining low level users. One spot (memcpy_real) was not touched,
though - fix it. Otherwise a kdump kernel will fail while reading
the old kernel. As we re-enable irqs with a non-standard function
we have to tell lockdep about that.

Fixes: 204ee2c5 ("s390/irqflags: optimize irq restore")
Signed-off-by: NChristian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

git commit dc7ee00d ("s390: lowcore stack pointer offsets")
introduced a regression in regard to s390_backtrace(). The stack
pointer for the asynchronous stack in the lowcore now has an
additional offset applied. This offset needs to be taken into account
in the calculation for the low and high address for the stack.

This bug was already partially fixed with commit 9cc5c206
("s390/dumpstack: fix address ranges for asynchronous and panic
stack"). This patch fixes it also for the oprofile code.

Fixes: dc7ee00d ("s390: lowcore stack pointer offsets")
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

git commit dc7ee00d ("s390: lowcore stack pointer offsets")
introduced a regression in regard to perf_callchain_kernel(). The
stack pointer for the asynchronous stack in the lowcore now has an
additional offset applied. This offset needs to be taken into account
in the calculation for the low and high address for the stack.

This bug was already partially fixed with 9cc5c206
("s390/dumpstack: fix address ranges for asynchronous and panic
stack"). This patch fixes it also for the perf_event code.

Fixes: dc7ee00d ("s390: lowcore stack pointer offsets")
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Implement save_stack_trace_regs, so that a stack trace of a kprobe
event can be obtained.

Without this we see following warning:
"save_stack_trace_regs() not implemented yet."
when we execute:
echo stacktrace > /sys/kernel/debug/tracing/trace_options
echo "p kfree" >> /sys/kernel/debug/tracing/kprobe_events
echo 1 > /sys/kernel/debug/tracing/events/kprobes/enable
Reported-by: NChunyu Hu <chuhu@redhat.com>
Signed-off-by: NPratyush Anand <panand@redhat.com>
[heiko.carstens@de.ibm.com]: changed patch to use __save_stack_trace()
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Tested-by: NPeter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

save_stack_trace() only saves the stack trace of the current context
(interrupt or process context). This is different to what other
architectures like x86 do, which save the full stack trace across
different contexts.

Also extract a __save_stack_trace() helper function which will be used
by a follow on patch.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Tested-by: NPeter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

save_stack_trace() did not write the ULONG_MAX end marker if there is
enough space left. So simply follow x86 and arm64.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Tested-by: NPeter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

git commit dc7ee00d ("s390: lowcore stack pointer offsets")
introduced a regression in regard to save_stack_trace(). The stack
pointer for the asynchronous and the panic stack in the lowcore now
have an additional offset applied to them. This offset needs to be
taken into account in the calculation for the low and high address for
the stacks.

This bug was already partially fixed with 9cc5c206
("s390/dumpstack: fix address ranges for asynchronous and panic
stack"). This patch fixes it also for the stacktrace code.

Fixes: dc7ee00d ("s390: lowcore stack pointer offsets")
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Tested-by: NPeter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The function save_stack_trace_tsk() did not consider that it can be
used for tsk == current, for which the current stack pointer obviously
cannot be found in the thread structure.

Fix this and get the stack pointer with an inline assembly.

This fixes e.g. the output of "cat /proc/self/stack".

Before:
[<0000000000000000>]           (null)
[<ffffffffffffffff>] 0xffffffffffffffff

After:
[<000000000011b3ee>] save_stack_trace_tsk+0x56/0x98
[<0000000000366cde>] proc_pid_stack+0xae/0x108
[<00000000003636f0>] proc_single_show+0x70/0xc0
[<0000000000311fbc>] seq_read+0xcc/0x448
[<00000000002e7716>] __vfs_read+0x36/0x100
[<00000000002e872e>] vfs_read+0x76/0x130
[<00000000002e975e>] SyS_read+0x66/0xd8
[<000000000089490e>] system_call+0xd6/0x264
[<ffffffffffffffff>] 0xffffffffffffffff
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Tested-by: NPeter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The kernel now always uses vector registers when available, however KVM
has special logic if support is really enabled for a guest. If support
is disabled, guest_fpregs.fregs will only contain memory for the fpu.
The kernel, however, will store vector registers into that area,
resulting in crazy memory overwrites.

Simply extending that area is not enough, because the format of the
registers also changes. We would have to do additional conversions, making
the code even more complex. Therefore let's directly use one place for
the vector/fpu registers + fpc (in kvm_run). We just have to convert the
data properly when accessing it. This makes current code much easier.

Please note that vector/fpu registers are now always stored to
vcpu->run->s.regs.vrs. Although this data is visible to QEMU and
used for migration, we only guarantee valid values to user space  when
KVM_SYNC_VRS is set. As that is only the case when we have vector
register support, we are on the safe side.

Fixes: b5510d9b ("s390/fpu: always enable the vector facility if it is available")
Cc: stable@vger.kernel.org # v4.4 d9a3a09a s390/kvm: remove dependency on struct save_area definition
Signed-off-by: NDavid Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: NChristian Borntraeger <borntraeger@de.ibm.com>
[adopt to d9a3a09a]

The KVM-VFIO device is used by the QEMU VFIO device. It is used to
record the list of in-use VFIO groups so that KVM can manipulate
them.
While we don't need this on s390 currently, let's try to be like
everyone else.
Signed-off-by: NDong Jia Shi <bjsdjshi@linux.vnet.ibm.com>
Acked-by: NCornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: NChristian Borntraeger <borntraeger@de.ibm.com>

fprs is never freed, therefore resulting in a memory leak if
kvm_vcpu_init() fails or the vcpu is destroyed.

Fixes: 9977e886 ("s390/kernel: lazy restore fpu registers")
Cc: stable@vger.kernel.org # v4.3+
Reported-by: NEric Farman <farman@linux.vnet.ibm.com>
Signed-off-by: NDavid Hildenbrand <dahi@linux.vnet.ibm.com>
Reviewed-by: NEric Farman <farman@linux.vnet.ibm.com>
Signed-off-by: NChristian Borntraeger <borntraeger@de.ibm.com>