The IDTE instruction used to flush TLB entries for a specific address
space uses the address-space-control element (ASCE) to identify
affected TLB entries. The upgrade of a page table adds a new top
level page table which changes the ASCE. The TLB entries associated
with the old ASCE need to be flushed and the ASCE for the address space
needs to be replaced synchronously on all CPUs which currently use it.
The concept of a lazy ASCE update with an exception handler is broken.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

There is a possible race between setting has_pgste and reallocation of the
page_table, change the order to fix this.
Also page_table_alloc_pgste can fail, in that case we need to backpropagte this
as -ENOMEM to the caller of page_table_realloc.

Based on a patch by Christian Borntraeger <borntraeger@de.ibm.com>.
Reviewed-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: NDominik Dingel <dingel@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Add a new Kconfig choice group which allows to configure how gcc should
tune the generated code (via -mtune option).
By default the -mtune parameter will match the -march parameter.

This is a rather large patch, but I wouldn't know how to make this shorter
unfortunately.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

this_cpu_xor() will be removed tree wide during the next merge window.
To avoid merge conflicts s390's removal comes via the s390 tree.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Use the ACCESS_ONCE macro for both accesses to idle->sequence in the
loops to calculate the idle time. If only one access uses the macro,
the compiler is free to cache the value for the second access which
can cause endless loops.

Cc: stable@vger.kernel.org # 3.6+
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The get_tod_clock_ext inline assembly does not specify its output
operands correctly. This can cause incorrect code to be generated.

Cc: stable@vger.kernel.org # 3.12
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

This typedef is unnecessary and should just be removed.
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>

Cleanup function information block used as a modify pci function
parameter. Change reserved members to be anonymous. Fix the size
of the struct and add proper alignment information. Also put the
FIB on the stack.
Signed-off-by: NSebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Add debugfs entries regardless of CONFIG_PCI_DEBUG.
Reviewed-by: NGerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: NSebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Cleanup arch specific pci messages. Remove unhelpful messages and
replace others with entries in the debugfs.
Reviewed-by: NGerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: NSebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Add four more defconfigs which we maintained outside of
the kernel tree for no good reason.
Inside of the kernel tree they can be easily used:

make whatever_defconfig

"make help" shows the available defconfigs.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Optimize this_cpu_* functions for 64 bit by making use of new instructions
that came with the interlocked-access facility 1 (load-and-*) and the
general-instructions-extension facility (asi, agsi).
That way we get rid of the compare-and-swap loop in most cases.
Code size reduction (defconfig, -march=z196): 11,555 bytes.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Remove the special cases for the this_cpu_* functions for 32 bit
in order to make it easier to add additional code for 64 bit.
32 bit will use the generic implementation.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Make psw32_user_bits a constant value again.
This is a leftover of the code which allowed to run the kernel either
in primary or home space which got removed with 9a905662 "s390/uaccess:
always run the kernel in home space".
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Fix the following bugs:
- When returning from a signal the signal handler copies the saved psw mask
  from user space and uses parts of it. Especially it restores the RI bit
  unconditionally. If however the machine doesn't support RI, or RI is
  disabled for the task, the last lpswe instruction which returns to user
  space will generate a specification exception.
  To fix this check if the RI bit is allowed to be set and kill the task
  if not.
- In the compat mode signal handler code the RI bit of the psw mask gets
  propagated to the mask of the return psw: if user space enables RI in the
  signal handler, RI will also be enabled after the signal handler is
  finished.
  This is a different behaviour than with 64 bit tasks. So change this to
  match the 64 bit semantics, which restores the original RI bit value.
- Fix similar oddities within the ptrace code as well.
Reviewed-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 FPC_VALID_MASK has been used to check the validity of the value
to be loaded into the floating-point-control register. With the
introduction of the floating-point extension facility and the
decimal-floating-point additional bits have been defined which need
to be checked in a non straight forward way. So far these bits have
been ignored which can cause an incorrect results for decimal-
floating-point operations, e.g. an incorrect rounding mode to be
set after signal return.

The static check with the FPC_VALID_MASK is replaced with a trial
load of the floating-point-control value, see test_fp_ctl.

In addition an information leak with the padding word between the
floating-point-control word and the floating-point registers in
the s390_fp_regs is fixed.
Reported-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Reviewed-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

If a machine has no hardware support for the xts-aes or ctr-aes algorithms
they are not registered in aes_s390_init. But aes_s390_fini unconditionally
unregisters the algorithms which causes crypto_remove_alg to crash.
Add two flag variables to remember if xts-aes and ctr-aes have been added.
Signed-off-by: NIngo Tuchscherer <ingo.tuchscherer@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Just change the type of "len" to unsigned int so the compiler can prove
that we don't have a buffer overflow (and generates less code).
We get rid of these:

In function 'copy_from_user',
    inlined from 'appldata_interval_handler' at
    arch/s390/appldata/appldata_base.c:265:
      uaccess.h:303: warning: call to 'copy_from_user_overflow' declared
      with attribute warning: copy_from_user() buffer size is not provably
      correct
In function 'copy_from_user',
    inlined from 'appldata_timer_handler' at
    arch/s390/appldata/appldata_base.c:225:
      uaccess.h:303: warning: call to 'copy_from_user_overflow' declared
      with attribute warning: copy_from_user() buffer size is not provably
      correct
In function 'copy_from_user',
    inlined from 'appldata_generic_handler' at
    arch/s390/appldata/appldata_base.c:333:
      uaccess.h:303: warning: call to 'copy_from_user_overflow' declared
      with attribute warning: copy_from_user() buffer size is not provably
      correct
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Get rid of these two warnings:

In function 'copy_from_user',
    inlined from 'cmm_timeout_handler' at arch/s390/mm/cmm.c:310:
      uaccess.h:303: warning: call to 'copy_from_user_overflow' declared
      with attribute warning: copy_from_user() buffer size is not provably
      correct
In function 'copy_from_user',
    inlined from 'cmm_pages_handler' at arch/s390/mm/cmm.c:270:
      uaccess.h:303: warning: call to 'copy_from_user_overflow' declared
      with attribute warning: copy_from_user() buffer size is not provably
      correct

Change the "len" type to unsigned int, so we can make sure that there is no
buffer overflow. This also generates less code.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Get rid of this one:

arch/s390/kernel/cache.c: In function 'cache_build_info':
arch/s390/kernel/cache.c:144: warning: 'private' may be used uninitialized
in this function
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Instead of returnin the number of bytes not copied and/or -EFAULT let the
signal handler helper functions always return -EFAULT if a user space
access failed.
This doesn't fix a bug in the current code, but makes is harder to get it
wrong in the future.
Also "smatch" won't complain anymore about the fact that the number of
remaining bytes gets returned instead of -EFAULT.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Signed-off-by: NMichael Holzheu <holzheu@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Missing parenthesis may cause problems when using the defines
together with operations of higher precedence.
Signed-off-by: NPeter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Currently zfpcdump can only collect registers for up to CONFIG_NR_CPUS
CPUss. This dependency is not necessary. So remove it by dynamically
allocating the save area array.
Signed-off-by: NMichael Holzheu <holzheu@linux.vnet.ibm.com>
Reviewed-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Steven Rostedt noted that s390 is the only architecture which calls
ftrace_push_return_trace() before ftrace_graph_entry() and therefore has
the small advantage that trace.depth gets initialized automatically.

However this small advantage isn't worth the difference and possible subtle
breakage that may result from this.
So change s390 to have the same function call order like all other
architectures: first ftrace_graph_entry(), then ftrace_push_return_trace()
Reported-by: NSteven Rostedt <rostedt@goodmis.org>
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Get rid of this compile warning:

arch/s390/kernel/crash_dump.c: In function 'copy_from_realmem':
arch/s390/kernel/crash_dump.c:48:6: warning: unused variable 'rc'
[-Wunused-variable]
  int rc;
      ^
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The type of 'v->counter' is always 'int', and related inline assembly
code also process 'int', so use 'unsigned int' instead of 'unsigned
long' for the 'mask'.
Signed-off-by: NChen Gang <gang.chen@asianux.com>
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

If [__]get_user_pages_fast() gets called with nr_pages == 0, the current
code would walk the page tables and pin as many pages until the first
invalid pte (or the kernel crashed while writing struct page pointers to
the pages array).
So let's handle at least the nr_pages == 0 case correctly and exit early.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Just call __get_user_pages_fast() from get_user_pages_fast() like powerpc.
This saves a lot of duplicated code.
Reviewed-by: NGerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

With dirty and referenced bits implemented in software it is unnecessary
to initialize the storage key for every page. With this patch not a single
storage key operation is done for a system that does not use KVM.
For KVM set_pte_at/pgste_set_key will do the initialization for the guest
view of the storage key when the mapping for the page is established in
the host.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

The prolog of functions generated by the bpf jit compiler uses an
instruction sequence with an "ahi" instruction to create stack space
instead of using an "aghi" instruction. Using the 32-bit "ahi" is not
wrong as the stack we are operating on is an order-4 allocation which
is always aligned to 16KB. But it is more consistent to use an "aghi"
as the stack pointer is a 64-bit value.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

- turn some macros into functions
- merge two almost identical versions for 32/64 bit
- add BUILD_BUG_ON() check to make sure the passed in array is large enough
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Implement architecture-specific functionality when a PCI device is
doing a hibernate transition.
Signed-off-by: NSebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Simplify the uaccess code by removing the user_mode=home option.
The kernel will now always run in the home space mode.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

find_first_bit_left() and friends have nothing to do with the normal
LSB0 bit numbering for big endian machines used in Linux (least
significant bit has bit number 0).
Instead they use MSB0 bit numbering, where the most signficant bit has
bit number 0. So rename find_first_bit_left() and friends to
find_first_bit_inv(), to avoid any confusion.
Also provide inv versions of set_bit, clear_bit and test_bit.

This also removes the confusing use of e.g. set_bit() in airq.c which
uses a "be_to_le" bit number conversion, which could imply that instead
set_bit_le() could be used. But that is entirely wrong since the _le
bitops variant uses yet another bit numbering scheme.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Since z9 109 we have the flogr instruction which can be used to implement
optimized versions of __ffs, ffs, __fls, fls and fls64.
So implement and use them, instead of the generic variants.
This reduces the size of the kernel image (defconfig, -march=z9-109)
by 19,648 bytes.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Just like all other architectures we should use out-of-line find bit
operations, since the inline variant bloat the size of the kernel image.
And also like all other architecures we should only supply optimized
variants of the __ffs, ffs, etc. primitives.

Therefore this patch removes the inlined s390 find bit functions and uses
the generic out-of-line variants instead.

The optimization of the primitives follows with the next patch.

With this patch also the functions find_first_bit_left() and
find_next_bit_left() have been reimplemented, since logically, they are
nothing else but a find_first_bit()/find_next_bit() implementation that
use an inverted __fls() instead of __ffs().
Also the restriction that these functions only work on machines which
support the "flogr" instruction is gone now.

This reduces the size of the kernel image (defconfig, -march=z9-109)
by 144,482 bytes.
Alone the size of the function build_sched_domains() gets reduced from
7 KB to 3,5 KB.

We also git rid of unused functions like find_first_bit_le()...
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Add the debug_level_enabled() function to check if debug events for
a particular level would be logged.  This might help to save cycles
for debug events that require additional information collection.
Signed-off-by: NHendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

Since zEC12 we have the interlocked-access facility 2 which allows to
use the instructions ni/oi/xi to update a single byte in storage with
compare-and-swap semantics.
So change set_bit(), clear_bit() and change_bit() to generate such code
instead of a compare-and-swap loop (or using the load-and-* instruction
family), if possible.
This reduces the text segment by yet another 8KB (defconfig).

Alternatively the long displacement variants niy/oiy/xiy could have
been used, but the extended displacement field is usually not needed
and therefore would only increase the size of the text segment again.
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>