On one of our testing machine, the following EINJ command lines:

  # echo 0x10000000 > param1
  # echo 0xfffffffffffff000 > param2
  # echo 0x8 > error_type
  # echo 1 > error_inject

Will get:

  echo: write error: Input/output error

The EIO comes from:

    rc = apei_exec_pre_map_gars(&trigger_ctx);

The root cause is as follow.  Normally, ACPI atomic IO support is used
to access IO memory.  But in EINJ of that machine, it is used to
access RAM to trigger the injected error.  And the ioremap() called by
apei_exec_pre_map_gars() can not map the RAM.

This patch add RAM mapping support to ACPI atomic IO support to
satisfy EINJ requirement.
Signed-off-by: NHuang Ying <ying.huang@intel.com>
Tested-by: NTony Luck <tony.luck@intel.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

There is no 64bit read/write support in ACPI atomicio because
readq/writeq is used to implement 64bit read/write, but readq/writeq
is not available on i386.  This patch implement 64bit read/write
support in atomicio via two readl/writel.
Signed-off-by: NHuang Ying <ying.huang@intel.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Callers to __acpi_ioremap_fast() pass the bit_width that they found in the
acpi_generic_address structure. Convert from bits to bytes when passing to
__acpi_find_iomap() - as it wants to see bytes, not bits.

cc: stable@kernel.org
Signed-off-by: NTony Luck <tony.luck@intel.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

If a file is only exporting symbols and not using the core
modular infrastructure, it can get by with just including
the smaller export.h header, which is a lot smaller than the
module.h header.
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

The presense of a writeq() implementation on 32-bit x86 that splits the
64-bit write into two 32-bit writes turns out to break the mpt2sas driver
(and in general is risky for drivers as was discussed in
<http://lkml.kernel.org/r/adaab6c1h7c.fsf@cisco.com>).  To fix this,
revert 2c5643b1 ("x86: provide readq()/writeq() on 32-bit too") and
follow-on cleanups.

This unfortunately leads to pushing non-atomic definitions of readq() and
write() to various x86-only drivers that in the meantime started using the
definitions in the x86 version of <asm/io.h>.  However as discussed
exhaustively, this is actually the right thing to do, because the right
way to split a 64-bit transaction is hardware dependent and therefore
belongs in the hardware driver (eg mpt2sas needs a spinlock to make sure
no other accesses occur in between the two halves of the access).

Build tested on 32- and 64-bit x86 allmodconfig.

Link: http://lkml.kernel.org/r/x86-32-writeq-is-broken@mdm.bga.comAcked-by: NHitoshi Mitake <h.mitake@gmail.com>
Cc: Kashyap Desai <Kashyap.Desai@lsi.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Ravi Anand <ravi.anand@qlogic.com>
Cc: Vikas Chaudhary <vikas.chaudhary@qlogic.com>
Cc: Matthew Garrett <mjg@redhat.com>
Cc: Jason Uhlenkott <juhlenko@akamai.com>
Acked-by: NJames Bottomley <James.Bottomley@parallels.com>
Acked-by: NIngo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

After we ioremap() a new region, we call __acpi_try_ioremap() to
see whether another thread has already mapped the same region.
This check clobbers "vaddr",  so compute the return value of
acpi_pre_map() using the ioremap() result "map->vaddr" instead.

v2:
    Modified the unsuitable description of patch.

v3:
    Removed unlikely() check and made description simpler.
Signed-off-by: NJin Dongming <jin.dongming@np.css.fujitsu.com>
Reviewed-by: NAndi Kleen <ak@linux.intel.com>
Signed-off-by: NHuang Ying <ying.huang@intel.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Implicit slab.h inclusion via percpu.h is about to go away.  Make sure
gfp.h or slab.h is included as necessary.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Len Brown <lenb@kernel.org>
Signed-off-by: NStephen Rothwell <sfr@canb.auug.org.au>

Some ACPI IO accessing need to be done in atomic context. For example,
APEI ERST operations may be used for permanent storage in hardware
error handler. That is, it may be called in atomic contexts such as
IRQ or NMI, etc. And, ERST/EINJ implement their operations via IO
memory/port accessing.  But the IO memory accessing method provided by
ACPI (acpi_read/acpi_write) maps the IO memory during it is accessed,
so it can not be used in atomic context. To solve the issue, the IO
memory should be pre-mapped during EINJ/ERST initializing. A linked
list is used to record which memory area has been mapped, when memory
is accessed in hardware error handler, search the linked list for the
mapped virtual address from the given physical address.
Signed-off-by: NHuang Ying <ying.huang@intel.com>
Signed-off-by: NAndi Kleen <ak@linux.intel.com>
Signed-off-by: NLen Brown <len.brown@intel.com>