The information in MCA records is filled in slightly differently on
Montecito than on Madison/McKinley.  Usually, the cache check and bus
check target identifiers have the same address.   On Montecito the
cache check and bus check target identifiers can be different if 
a corrected error (ie SBE or unconsumed poison data) was encountered and
then an uncorrected error (ie DBE) was consumed.  In that case, the 
cache check target identifier is the physical address of the DBE (that
caused the MCA to surface) while the bus check target identifier is the 
physical address of the SBE.  This patch correctly finds the target
identifier that triggered the MCA.

If there are multiple valid cache target identifiers in the same
error record then use the one with the lowest cache level.

Signed-off-by: Russ Anderson (rja@sgi.com)
Signed-off-by: NTony Luck <tony.luck@intel.com>

Printing message to console from MCA/INIT handler is useful,
however doing oops_in_progress = 1 in them exactly makes
something in kernel wrong. Especially it sounds ugly if
system goes wrong after returning from recoverable MCA.

This patch adds ia64_mca_printk() function that collects
messages into temporary-not-so-large message buffer during
in MCA/INIT environment and print them out later, after
returning to normal context or when handlers determine to
down the system.

Also this print function is exported for use in extensional
MCA handler. It would be useful to describe detail about
recovery.

NOTE:
I don't think it is sane thing if temporary message buffer
is enlarged enough to hold whole stack dumps from INIT, so
buffering is disabled during stack dump from INIT-monarch
(= default_monarch_init_process). please fix it in future.
Signed-off-by: NHidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Acked-by: NRuss Anderson <rja@sgi.com>
Signed-off-by: NTony Luck <tony.luck@intel.com>

Signed-off-by: NJörn Engel <joern@wohnheim.fh-wedel.de>
Signed-off-by: NAdrian Bunk <bunk@stusta.de>

When the mca recovery code encounters a condition that makes
the MCA non-recoverable, print the reason it could not recover.
This will make it easier to identify why the recovery code did
not recover.
Signed-off-by: NRuss Anderson <rja@sgi.com>
Signed-off-by: NTony Luck <tony.luck@intel.com>

Memory errors encountered by user applications may surface
when the CPU is running in kernel context.  The current code
will not attempt recovery if the MCA surfaces in kernel
context (privilage mode 0).  This patch adds a check for cases
where the user initiated the load that surfaces in kernel
interrupt code.

An example is a user process lauching a load from memory
and the data in memory had bad ECC.  Before the bad data
gets to the CPU register, and interrupt comes in.  The
code jumps to the IVT interrupt entry point and begins
execution in kernel context.  The process of saving the
user registers (SAVE_REST) causes the bad data to be loaded
into a CPU register, triggering the MCA.  The MCA surfaces in
kernel context, even though the load was initiated from
user context.

As suggested by David and Tony, this patch uses an exception
table like approach, puting the tagged recovery addresses in
a searchable table.  One difference from the exception table
is that MCAs do not surface in precise places (such as with
a TLB miss), so instead of tagging specific instructions,
address ranges are registers.  A single macro is used to do
the tagging, with the input parameter being the label
of the starting address and the macro being the ending
address.  This limits clutter in the code.

This patch only tags one spot, the interrupt ivt entry.
Testing showed that spot to be a "heavy hitter" with
MCAs surfacing while saving user registers.  Other spots
can be added as needed by adding a single macro.

Signed-off-by: Russ Anderson (rja@sgi.com)
Signed-off-by: NTony Luck <tony.luck@intel.com>

When there is no bus check, the return code should be failure, not success.

Signed-off-by: Russ Anderson (rja@sgi.com)
Signed-off-by: NTony Luck <tony.luck@intel.com>

The MCA recovery messages are currently KERN_DEBUG,
so they don't show up in /var/log/messages (by default).
Increase the severity to KERN_ERR, for the initial
message (and also add the physical address to this
message). Leave the successful isolation message as
KERN_DEBUG, but increase the severity when isolation
fails to KERN_CRIT.

[Russ' patch made these all KERN_CRIT]

Signed-off-by: Russ Anderson (rja@sgi.com)
Signed-off-by: NTony Luck <tony.luck@intel.com>

MCA driver can cause panic if kernel gets a state info with no minstate.
This patch adds minstate validation before handling it.
Signed-off-by: NHidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: NTony Luck <tony.luck@intel.com>

When a page has a memory uncorrectable ECC error, the recovery
code wants to prevent the page from being reused.  This change
bumps the reference count to prevent the page from getting back
on the free list.

Signed-off-by: Russ Anderson (rja@sgi.com)
Signed-off-by: NTony Luck <tony.luck@intel.com>

paddr needs to be shifted by PAGE_SHIFT to be valid
input for pfn_valid().
Signed-off-by: NRuss Anderson <rja@sgi.com>
Signed-off-by: NTony Luck <tony.luck@intel.com>

The determination of whether an MCA is recoverable or not must
be based on the bits set in the PSP (Processor State Parameter).
The specific bits are shown in the Intel IA-64 Architecture Software
Developer's Manual, Vol 2, Table 11-6 Software Recovery Bits in
Processor State Parameter.  Those bits should be consistent
across the entire IA-64 family of processors.
Signed-off-by: NRuss Anderson <rja@sgi.com>
Signed-off-by: NTony Luck <tony.luck@intel.com>

Verify the pfn is valid before calling pfn_to_page(),
and cut isolation message if nothing was done.
Signed-off-by: NHidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Acked-by: NRuss Anderson <rja@sgi.com>
Signed-off-by: NTony Luck <tony.luck@intel.com>

There were some trailing white spaces, long lines, brackets in
weird style etc.  This patch cleans them up.
Signed-off-by: NHidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: NTony Luck <tony.luck@intel.com>

The bulk of the change.  Use per cpu MCA/INIT stacks.  Change the SAL
to OS state (sos) to be per process.  Do all the assembler work on the
MCA/INIT stacks, leaving the original stack alone.  Pass per cpu state
data to the C handlers for MCA and INIT, which also means changing the
mca_drv interfaces slightly.  Lots of verification on whether the
original stack is usable before converting it to a sleeping process.
Signed-off-by: NKeith Owens <kaos@sgi.com>
Signed-off-by: NTony Luck <tony.luck@intel.com>

Jack Steiner uncovered some opportunities for improvement in
the MCA recovery code.

  1) Set bsp to save registers on the kernel stack.
  2) Disable interrupts while in the MCA recovery code.
  3) Change the way the user process is killed, to avoid 
     a panic in schedule.

Testing shows that these changes make the recovery code much 
more reliable with the 2.6.12 kernel.
Signed-off-by: NRuss Anderson <rja@sgi.com>
Signed-off-by: NTony Luck <tony.luck@intel.com>

Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.

Let it rip!