Impact: standardize IO on cached ops

On modern CPUs it is almost always a bad idea to use non-temporal stores,
as the regression in this commit has shown it:

  30d697fa: x86: fix performance regression in write() syscall

The kernel simply has no good information about whether using non-temporal
stores is a good idea or not - and trying to add heuristics only increases
complexity and inserts fragility.

The regression on cached write()s took very long to be found - over two
years. So dont take any chances and let the hardware decide how it makes
use of its caches.

The only exception is drivers/gpu/drm/i915/i915_gem.c: there were we are
absolutely sure that another entity (the GPU) will pick up the dirty
data immediately and that the CPU will not touch that data before the
GPU will.

Also, keep the _nocache() primitives to make it easier for people to
experiment with these details. There may be more clear-cut cases where
non-cached copies can be used, outside of filemap.c.

Cc: Salman Qazi <sqazi@google.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: make more types of copies non-temporal

This change makes the following simple fix:

  30d697fa: x86: fix performance regression in write() syscall

A bit more sophisticated: we check the 'total' number of bytes
written to decide whether to copy in a cached or a non-temporal
way.

This will for example cause the tail (modulo 4096 bytes) chunk
of a large write() to be non-temporal too - not just the page-sized
chunks.

Cc: Salman Qazi <sqazi@google.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup, enable future change

Add a 'total bytes copied' parameter to __copy_from_user_*nocache(),
and update all the callsites.

The parameter is not used yet - architecture code can use it to
more intelligently decide whether the copy should be cached or
non-temporal.

Cc: Salman Qazi <sqazi@google.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup

one 32-bit system reports:

BIOS-provided physical RAM map:
 BIOS-e820: 0000000000000000 - 000000000009fc00 (usable)
 BIOS-e820: 000000000009fc00 - 00000000000a0000 (reserved)
 BIOS-e820: 00000000000f0000 - 0000000000100000 (reserved)
 BIOS-e820: 0000000000100000 - 000000001c000000 (usable)
 BIOS-e820: 00000000ffff0000 - 0000000100000000 (reserved)
DMI 2.0 present.
last_pfn = 0x1c000 max_arch_pfn = 0x100000
kernel direct mapping tables up to 1c000000 @ 7000-c000
..
RAMDISK: 1bc69000 - 1bfef4fa
..
0MB HIGHMEM available.
448MB LOWMEM available.
  mapped low ram: 0 - 1c000000
  low ram: 00000000 - 1c000000
  bootmap 00002000 - 00005800
(9 early reservations) ==> bootmem [0000000000 - 001c000000]
  #0 [0000000000 - 0000001000]   BIOS data page ==> [0000000000 - 0000001000]
  #1 [0000001000 - 0000002000]    EX TRAMPOLINE ==> [0000001000 - 0000002000]
  #2 [0000006000 - 0000007000]       TRAMPOLINE ==> [0000006000 - 0000007000]
  #3 [0000400000 - 00009ed14c]    TEXT DATA BSS ==> [0000400000 - 00009ed14c]
  #4 [001bc69000 - 001bfef4fa]          RAMDISK ==> [001bc69000 - 001bfef4fa]
  #5 [00009ee000 - 00009f2000]    INIT_PG_TABLE ==> [00009ee000 - 00009f2000]
  #6 [000009f400 - 0000100000]    BIOS reserved ==> [000009f400 - 0000100000]
  #7 [0000007000 - 0000007000]          PGTABLE
  #8 [0000002000 - 0000006000]          BOOTMAP ==> [0000002000 - 0000006000]

Notice the strange blank PGTABLE entry.

The reason is init_pg_table is big enough, and zero range is called
with init_memory_mapping/reserve_early().

So try to check the range in reserve_early()

v2: fix the reversed compare
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: nickpiggin@yahoo.com.au
Cc: ink@jurassic.park.msu.ru
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: heukelum@fastmail.fm
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: heukelum@fastmail.fm
Signed-off-by: Ingo Molnar <mingo@elte.hu>

native_usergs_sysret64 is described as

	extern void native_usergs_sysret64(void)

so lets add ENDPROC here
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: heukelum@fastmail.fm
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: heukelum@fastmail.fm
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: heukelum@fastmail.fm
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup

NEXT_PAGE already has 'balign' so no
need to keep this redundant one.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: heukelum@fastmail.fm
Signed-off-by: Ingo Molnar <mingo@elte.hu>

While the introduction of __copy_from_user_nocache (see commit:
0812a579) may have been an improvement
for sufficiently large writes, there is evidence to show that it is
deterimental for small writes.  Unixbench's fstime test gives the
following results for 256 byte writes with MAX_BLOCK of 2000:

    2.6.29-rc6 ( 5 samples, each in KB/sec ):
    283750, 295200, 294500, 293000, 293300

    2.6.29-rc6 + this patch (5 samples, each in KB/sec):
    313050, 3106750, 293350, 306300, 307900

    2.6.18
    395700, 342000, 399100, 366050, 359850

    See w_test() in src/fstime.c in unixbench version 4.1.0.  Basically, the above test
    consists of counting how much we can write in this manner:

    alarm(10);
    while (!sigalarm) {
            for (f_blocks = 0; f_blocks < 2000; ++f_blocks) {
                   write(f, buf, 256);
            }
            lseek(f, 0L, 0);
    }

Note, there are other components to the write syscall regression
that are not addressed here.
Signed-off-by: Salman Qazi <sqazi@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Clarify the kmmio_fault() comment.
Acked-by: Pekka Paalanen <pq@iki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: extend prefetch handling on 64-bit

Currently there's an extra is_prefetch() check done in do_sigbus(),
which we only do on 32 bits.

This is a last-ditch check before we terminate a task, so it's worth
giving prefetch instructions another chance - should none of our
existing quirks have caught a prefetch instruction related spurious
fault.

The only risk is if a prefetch causes a real sigbus, in that case
we'll not OOM but try another fault. But this code has been on
32-bit for a long time, so it should be fine in practice.

So do this on 64-bit too - and thus remove one more #ifdef.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup

Removal of an #ifdef in fault_in_kernel_space(), by making
use of the new TASK_SIZE_MAX symbol which is now available
on 32-bit too.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup

Rename TASK_SIZE64 to TASK_SIZE_MAX, and provide the
define on 32-bit too. (mapped to TASK_SIZE)

This allows 32-bit code to make use of the (former-) TASK_SIZE64
symbol as well, in a clean way.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup

do_page_fault() has this ugly #ifdef in its prototype:

  #ifdef CONFIG_X86_64
  asmlinkage
  #endif
  void __kprobes do_page_fault(struct pt_regs *regs, unsigned long error_code)

Replace it with 'dotraplinkage' which maps to exactly the above
construct: nothing on 32-bit and asmlinkage on 64-bit.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: add oops-recursion check to 32-bit

Unify the oops state-machine, to the 64-bit version. It is
slightly more careful in that it does a recursion check
in oops_begin(), and is thus more likely to show the relevant
oops.

It also means that 32-bit will print one more line at the
end of pagefault triggered oopses:

 	printk(KERN_EMERG "CR2: %016lx\n", address);

Which is generally good information to be seen in partial-dump
digital-camera jpegs ;-)

The downside is the somewhat more complex critical path. Both
variants have been tested well meanwhile by kernel developers
crashing their boxes so i dont think this is a practical worry.

This removes 3 ugly #ifdefs from no_context() and makes the
function a lot nicer read.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: refine/extend page fault related oops printing on 64-bit

 - honor the pause_on_oops logic on 64-bit too
 - print out NX fault warnings on 64-bit as well
 - factor out the NX fault message to make it git-greppable and readable

Note that this means that we do the PF_INSTR check on 32-bit non-PAE
as well where it should not occur ... normally. Cannot hurt.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup

Avoid a couple more #ifdefs by moving fundamentally non-unifiable
functions into a single #ifdef 32-bit / #else / #endif block in
fault.c: vmalloc*(), dump_pagetable(), check_vm8086_mode().

No code changed:

   text	   data	    bss	    dec	    hex	filename
   4618	     32	     24	   4674	   1242	fault.o.before
   4618	     32	     24	   4674	   1242	fault.o.after

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup

Remove an #ifdef from notify_page_fault(). The function still
compiles to nothing in the !CONFIG_KPROBES case.

Introduce kprobes_built_in() and kprobe_fault_handler() helpers
to allow this - they returns 0 if !CONFIG_KPROBES.

No code changed:

   text	   data	    bss	    dec	    hex	filename
   4618	     32	     24	   4674	   1242	fault.o.before
   4618	     32	     24	   4674	   1242	fault.o.after

Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup

Remove an #ifdef from kmmio_fault() - we can do this by
providing default implementations for is_kmmio_active()
and kmmio_handler(). The compiler optimizes it all away
in the !CONFIG_MMIOTRACE case.

Also, while at it, clean up mmiotrace.h a bit:

 - standard header guards
 - standard vertical spaces for structure definitions

No code changed (both with mmiotrace on and off in the config):

   text	   data	    bss	    dec	    hex	filename
   2947	     12	     12	   2971	    b9b	fault.o.before
   2947	     12	     12	   2971	    b9b	fault.o.after

Cc: Pekka Paalanen <pq@iki.fi>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: improve page fault handling robustness

The 'PF_RSVD' flag (bit 3) of the page-fault error_code is a
relatively recent addition to x86 CPUs, so the 32-bit do_fault()
implementation never had it. This flag gets set when the CPU
detects nonzero values in any reserved bits of the page directory
entries.

Extend the existing 64-bit check for PF_RSVD in do_page_fault()
to 32-bit too. If we detect such a fault then we print a more
informative oops and the pagetables.

This unifies the code some more, removes an ugly #ifdef and improves
the 32-bit page fault code robustness a bit. It slightly increases
the 32-bit kernel text size.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup

Instead of an ugly, open-coded, #ifdef-ed vm86 related legacy check
in do_page_fault(), put it into the check_v8086_mode() helper
function and merge it with an existing #ifdef.

Also, simplify the code flow a tiny bit in the helper.

No code changed:

arch/x86/mm/fault.o:

   text	   data	    bss	    dec	    hex	filename
   2711	     12	     12	   2735	    aaf	fault.o.before
   2711	     12	     12	   2735	    aaf	fault.o.after

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: no functionality changed

Factor out the opcode checker into a helper inline.

The code got a tiny bit smaller:

   text	   data	    bss	    dec	    hex	filename
   4632	     32	     24	   4688	   1250	fault.o.before
   4618	     32	     24	   4674	   1242	fault.o.after

And it got cleaner / easier to review as well.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup, no code changed

Clean up various small details, which can be correctness checked
automatically:

 - tidy up the include file section
 - eliminate unnecessary includes
 - introduce show_signal_msg() to clean up code flow
 - standardize the code flow
 - standardize comments and other style details
 - more cleanups, pointed out by checkpatch

No code changed on either 32-bit nor 64-bit:

arch/x86/mm/fault.o:

   text	   data	    bss	    dec	    hex	filename
   4632	     32	     24	   4688	   1250	fault.o.before
   4632	     32	     24	   4688	   1250	fault.o.after

the md5 changed due to a change in a single instruction:

   2e8a8241e7f0d69706776a5a26c90bc0  fault.o.before.asm
   c5c3d36e725586eb74f0e10692f0193e  fault.o.after.asm

Because a __LINE__ reference in a WARN_ONCE() has changed.

On 32-bit a few stack offsets changed - no code size difference
nor any functionality difference.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge branch 'tip/x86/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace into x86/mm

Impact: future-proof the split_large_page() function

Linus noticed that split_large_page() is not safe wrt. the
PAT bit: it is bit 12 on the 1GB and 2MB page table level
(_PAGE_BIT_PAT_LARGE), and it is bit 7 on the 4K page
table level (_PAGE_BIT_PAT).

Currently it is not a problem because we never set
_PAGE_BIT_PAT_LARGE on any of the large-page mappings - but
should this happen in the future the split_large_page() would
silently lift bit 12 into the lowlevel 4K pte and would start
corrupting the physical page frame offset. Not fun.

So add a debug warning, to make sure if something ever sets
the PAT bit then this function gets updated too.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: fix to prevent hard lockup on bad PMD permissions

If the PMD does not have the correct permissions for a page access,
but the PTE does, the spurious fault handler will mistake the fault
as a lazy TLB transaction. This will result in an infinite loop of:

 fault -> spurious_fault check (pass) -> return to code -> fault

This patch adds a check and a warn on if the PTE passes the permissions
but the PMD does not.

[ Updated: Ingo Molnar suggested using WARN_ONCE with some text ]
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

Impact: Cleanup. No functional changes.
Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Steven Rostedt found a bug in where in his modified kernel
ftrace was unable to modify the kernel text, due to the PMD
itself having been marked read-only as well in
split_large_page().

The fix, suggested by Linus, is to not try to 'clone' the
reference protection of a huge-page, but to use the standard
(and permissive) page protection bits of KERNPG_TABLE.

The 'cloning' makes sense for the ptes but it's a confused and
incorrect concept at the page table level - because the
pagetable entry is a set of all ptes and hence cannot
'clone' any single protection attribute - the ptes can be any
mixture of protections.

With the permissive KERNPG_TABLE, even if the pte protections
get changed after this point (due to ftrace doing code-patching
or other similar activities like kprobes), the resulting combined
protections will still be correct and the pte's restrictive
(or permissive) protections will control it.

Also update the comment.

This bug was there for a long time but has not caused visible
problems before as it needs a rather large read-only area to
trigger. Steve possibly hacked his kernel with some really
large arrays or so. Anyway, the bug is definitely worth fixing.

[ Huang Ying also experienced problems in this area when writing
  the EFI code, but the real bug in split_large_page() was not
  realized back then. ]
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Reported-by: Huang Ying <ying.huang@intel.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: fix time warps under vmware

Similar to the check for TSC going backwards in the TSC clocksource,
we also need this check for VMI clocksource.
Signed-off-by: Alok N Kataria <akataria@vmware.com>
Cc: Zachary Amsden <zach@vmware.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: stable@kernel.org

* master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] 5405/1: ep93xx: remove unused gesbc9312.h header
  [ARM] 5404/1: Fix condition in arm_elf_read_implies_exec() to set READ_IMPLIES_EXEC
  [ARM] omap: fix clock reparenting in omap2_clk_set_parent()
  [ARM] 5403/1: pxa25x_ep_fifo_flush() *ep->reg_udccs always set to 0
  [ARM] 5402/1: fix a case of wrap-around in sanity_check_meminfo()
  [ARM] 5401/1: Orion: fix edge triggered GPIO interrupt support
  [ARM] 5400/1: Add support for inverted rdy_busy pin for Atmel nand device controller
  [ARM] 5391/1: AT91: Enable GPIO clocks earlier
  [ARM] 5390/1: AT91: Watchdog fixes
  [ARM] 5398/1: Add Wan ZongShun to MAINTAINERS for W90P910
  [ARM] omap: fix _omap2_clksel_get_src_field()
  [ARM] omap: fix omap2_divisor_to_clksel() error return value

Merge branch 'x86/untangle2' of git://git.kernel.org/pub/scm/linux/kernel/git/jeremy/xen into x86/headers

Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, mce: fix ifdef for 64bit thermal apic vector clear on shutdown
  x86, mce: use force_sig_info to kill process in machine check
  x86, mce: reinitialize per cpu features on resume
  x86, rcu: fix strange load average and ksoftirqd behavior

Merge branch 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  tracing: limit the number of loops the ring buffer self test can make
  tracing: have function trace select kallsyms
  tracing: disable tracing while testing ring buffer
  tracing/function-graph-tracer: trace the idle tasks