Allow csrows to properly initialize when the topology only has active
channels on 2 and 3. This new check allows proper detection and
initialization in this topology. Only checking the first mrt that
represented channels 0 and 1 is not sufficient.
I also fixed up the related debug information path. I can submit as a 2nd
patch if needed.
Signed-off-by: Keith Mannthey <kmannth@us.ibm.com>
Acked-by: Aristeu Rozanski <aris@ruivo.org>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When building without CONFIG_PCI the edac_pci_idx variable is unused,
causing a build-time warning. Wrap the variable in #ifdef CONFIG_PCI,
just like the rest of the PCI support.
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The i5400 EDAC driver has several bugs with chip-select row computation
which most likely lead to bugs in detailed error reporting. Attempts to
contact the authors have gone mostly unanswered so I am presenting my diff
here. I do not subscribe to lkml and would appreciate being kept in the
cc.
The most egregious problem was miscalculating the addresses of MTR
registers after register 0 by assuming they are 32bit rather than 16.
This caused the driver to miss half of the memories. Most motherboards
tend to have only 8 dimm slots and not 16, so this may not have been
noticed before.
Further, the row calculations multiplied the number of dimms several
times, ultimately ending up with a maximum row of 32. The chipset only
supports 4 dimms in each of 4 channels, so csrow could not be higher than
4 unless you use a row per-rank with dual-rank dimms. I opted to
eliminate this behavior as it is confusing to the user and the error
reporting works by slot and not rank. This gives a much clearer view of
memory by slot and channel in /sys.
Signed-off-by: Jeff Roberson <jroberson@jroberson.net>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When injecting DRAM ECC errors (F3xBC_x8), EccVector[15:0] is a bitmask
of which bits should be error injected when written to and holds the
payload of 16-bit DRAM word when read, respectively.
Add /sysfs members to show the DRAM ECC section/word/vector.
Fail wrong injection values entered over /sysfs instead of truncating
them.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
On Fam10h and above, F1x[1, 0][7C:40] are DRAM Base/Limit registers
which specify the destination node of a DRAM address. Those address
boundaries are being extracted into ->dram_base[] and ->dram_limit[].
Correct the extraction masks to match the respective address bits.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Different processor families support a different number of chip selects.
Handle this in a family-dependent way with the proper values assigned at
init time (see amd64_set_dct_base_and_mask).
Remove _DCSM_COUNT defines since they're used at one place and originate
from public documentation.
CC: Keith Mannthey <kmannth@us.ibm.com>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
This allows the errors to be further decoded and mapped to csrows.
Tested with ECC debug dimms and an Rev F cpu based system.
Signed-off-by: Keith Mannthey <kmannth@us.ibm.com>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
The check when DRAM interleaving is enabled should be done against the
pvt->dram_IntlvSel field and not against the ->dram_limit.
Simplify first loop and fixup printk formatting while at it.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
The pvt->dram_IntlvEn saves the 3 "Interleave Enable" bits already
right-shifted by 8 so the check in find_mc_by_sys_addr() by shifting the
values to the left 8 bits is wrong.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
K8 DRAM base and limit addresses from F1x40 +8*i and F1x44 + 8*i, where
i in (0..7) are both bits 39-24 and therefore the shifting should be
done by 24 and not by 8.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Allocate memory statically for 8-node machines max for simplicity
instead of relying on MAX_NUMNODES which is 0 on !CONFIG_NUMA builds.
Spotted by Jan Beulich.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
This converts the MCE decoding logic into a standalone config
option which can be built-in or a module, the first one being the
default for MCEs happening early on in the boot process.
This, beyond being separated in a cleaner way, also saves RAM by
making the decoding logic modular.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andi Kleen <andi@firstfloor.org>
LKML-Reference: <20091002133148.GD28682@aftab>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Make decoding of MCEs happen only on AMD hardware by registering a
non-default callback only on CPU families which support it.
While looking at the interaction of decode_mce() with the other MCE
code i also noticed a few other things and made the following
cleanups/fixes:
- Fixed the mce_decode() weak alias - a weak alias is really not
good here, it should be a proper callback. A weak alias will be
overriden if a piece of code is built into the kernel - not
good, obviously.
- The patch initializes the callback on AMD family 10h and 11h.
- Added the more correct fallback printk of:
No support for human readable MCE decoding on this CPU type.
Transcribe the message and run it through 'mcelog --ascii' to decode.
On CPUs that dont have a decoder.
- Made the surrounding code more readable.
Note that the callback allows us to have a default fallback -
without having to check the CPU versions during the printout
itself. When an EDAC module registers itself, it can install the
decode-print function.
(there's no unregister needed as this is core code.)
version -v2 by Borislav Petkov:
- add K8 to the set of supported CPUs
- always build in edac_mce_amd since we use an early_initcall now
- fix checkpatch warnings
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andi Kleen <andi@firstfloor.org>
LKML-Reference: <20091001141432.GA11410@aftab>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Module edac_core.ko uses call_rcu() callbacks in edac_device.c, edac_mc.c
and edac_pci.c.
They all use a wait_for_completion() scheme, but this scheme it not 100%
safe on multiple CPUs. See the _rcu_barrier() implementation which
explains why extra precausion is needed.
The patch adds a comment about rcu_barrier() and as a precausion calls
rcu_barrier(). A maintainer needs to look at removing the
wait_for_completion code.
[dougthompson@xmission.com: remove the wait_for_completion code]
Signed-off-by Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A driver for the Intel 3200 and 3210 memory controllers. It has only had
light testing so far, and currently makes no attempt to decode error
addresses at anything finer than csrow granularity.
Signed-off-by: Jason Uhlenkott <juhlenko@akamai.com>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use the function resource_size, which reduces the chance of introducing
off-by-one errors in calculating the resource size.
The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)
// <smpl>
@@
struct resource *res;
@@
- (res->end - res->start) + 1
+ resource_size(res)
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add support for the Freescale MPC83xx memory controller to the existing
driver for the Freescale MPC85xx memory controller. The only difference
between the two processors are in the CS_BNDS register parsing code, which
has been changed so it will work on both processors.
The L2 cache controller does not exist on the MPC83xx, but the OF
subsystem will not use the driver if the device is not present in the OF
device tree.
I had to change the nr_pages calculation to make the math work out. I
checked it on my board and did the math by hand for a 64GB 85xx using 64K
pages. In both cases, nr_pages * PAGE_SIZE comes out to the correct
value.
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Cc: Kumar Gala <galak@gate.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Based on Kumar's new compatible types patch, add P2020 into MPC85xx EDAC
compatible lists so that EDAC can recognize P2020 meomry controller and L2
cache controller and export the relevant fields to sysfs.
EDAC MPC85xx DDR3 support is needed if DDR3 memory stick is installed on a
P2020DS board so that EDAC core can recognize DDR3 memory type.
Signed-off-by: Yang Shi <yang.shi@windriver.com>
Acked-by: Dave Jiang <djiang@mvista.com>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The old code was using smp_call_function_many which skips the current
cpu if it is in the supplied cpumask. Switch to the rdmsr_on_cpus()
interface which takes care of that.
In addition, add get_cpus_on_this_dct_cpumask helper which computes a
cpumask of all the cores on a node and thus on a DCT.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Simplify the procedure by checking if there is any DIMM in each channel.
This patch will fix the bugs such as when there is no DIMMs under
certain node, two DIMMs in the same channel, and only one DIMM in each
channel of the node.
Borislav: minor fixups
Signed-off-by: Wan Wei <wanwei@mail.dawning.com.cn>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Simplify code flow and make sure return value is always valid since
further driver init depends on it. Carve out long warning string and
make code more readable. Shorten some names, while at it.
There should be no functional change resulting from this patch.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Acked-by: H. Peter Anvin <hpa@zytor.com>
-tip testing found the following build failure (config attached):
drivers/built-in.o: In function `amd64_check':
amd64_edac.c:(.text+0x3e9491): undefined reference to `amd_decode_nb_mce'
drivers/built-in.o: In function `amd64_init_2nd_stage':
amd64_edac.c:(.text+0x3e9b46): undefined reference to `amd_report_gart_errors'
amd64_edac.c:(.text+0x3e9b55): undefined reference to `amd_register_ecc_decoder'
drivers/built-in.o: In function `amd64_nbea_store':
amd64_edac_dbg.c:(.text+0x3ea22e): undefined reference to `amd_decode_nb_mce'
drivers/built-in.o: In function `amd64_remove_one_instance':
amd64_edac.c:(.devexit.text+0x3eea): undefined reference to `amd_report_gart_errors'
amd64_edac.c:(.devexit.text+0x3ef6): undefined reference to `amd_unregister_ecc_decoder'
the AMD EDAC code has a dependency on CONFIG_CPU_SUP_AMD facilities. The
patch below solves the problem here.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Those get reported in MC0_STATUS, see Table 92, F10h BKDG (31116, rev.
3.28) for more details.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
This is the MCE error code from the MCi_STATUS banks, bits [15:0] which
describe what type of error was encountered: GART TLB, Memory or Bus
error. The semantics of those bits are identical across all MCE banks so
decode those separately, irrespectively of MCE type.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
The MCi_STATUS registers have most field definitions in common so decode
them in the general path. Do not pass ecc_type along and compute it in
__amd64_decode_bus_error instead.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Move NB decoder along with required defines to EDAC MCE core. Add
registration routines for further decoding of the MCE info in the AMD64
EDAC module.
CC: Andi Kleen <andi@firstfloor.org>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
This is in preparation of adding AMD-specific MCE decoding functionality
to the EDAC core. The error decoding macros originate from the AMD64
EDAC driver albeit in a simplified and cleaned up version here.
While at it, add macros to generate the error description strings and
use them in the error type decoders directly which removes a bunch of
code and makes the decoding functions much more readable. Also, fix
strings and shorten macro names.
Remove superfluous htlink_msgs.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
On the good path of BIOS enabled ECC and no override, the value returned
is 1 by omission and thus is deemed failing by the probe-function.
Allow proper module initialization by clearing the retval explicitly.
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Intel X38 MCHBAR is a 64bits register, base from 0x48, so its higher base
is 0x4C.
Signed-off-by: Lu Zhihe <tombowfly@gmail.com>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Cc: <stable@kernel.org> [2.6.30.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Since some new MPC85xx SOCs support DDR3 memory now, so add DDR3 memory
type for MPC85xx EDAC.
Signed-off-by: Yang Shi <yang.shi@windriver.com>
Cc: Doug Thompson <norsk5@yahoo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
amd64_check_ecc_enabled() returns non-zero status when ECC
checking/correcting is disabled and this fails further loading of the
driver even when 'ecc_enable_override' boot param is used.
Fix that by clearing return status in that case.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Checking whether the machine is using ECC enabled DRAM is done through
testing the DimmEccEn bit in the DRAM Cfg Low register (F2x[1,0]90). Do
that instead of testing all bits from the DimmEccEn upwards.
Also, remove mci->edac_cap assignment and use value returned from
amd64_determine_edac_cap().
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
The remove function uses __devexit, so the .remove assignment needs
__devexit_p() to fix a build error with hotplug disabled.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Cc: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Keith Mannthey