Xiaolong Ye reported the following failure on Broadwell D server:
EDAC sbridge: Some needed devices are missing
EDAC MC: Removed device 0 for sbridge_edac.c Broadwell SrcID#0_Ha#0: DEV 0000:ff:12.0
EDAC sbridge: Couldn't find mci handler
EDAC sbridge: Failed to register device with error -19.
Broadwell D (only IMC0 per socket) and Broadwell X (IMC0 and IMC1 per
socket) use the same PCI device IDs for IMC0 per socket, then they
share pci_dev_descr_broadwell_table (n_imcs_per_sock=2). In this case,
Broadwell D wrongly creates the nonexistent SOCK EDAC memory controller
and reports above error messages, since it has no IMC1 per socket.
Avoid creating the nonexistent SOCK memory controller.
Reported-and-tested-by: Xiaolong Ye <xiaolong.ye@intel.com>
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20170608113351.25323-1-qiuxu.zhuo@intel.com
[ Massage. ]
Signed-off-by: Borislav Petkov <bp@suse.de>
Collapse 'case:' in *_mci_bind_devs() and update driver version from
1.1.1 to 1.1.2.
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20170523000934.87971-1-qiuxu.zhuo@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
This is based on previous work by Patrick Geary, see Link.
Additional cleanups ontop:
- Remove the code to read MCMTR from pci_ha1_ta and CHN_TO_HA macro,
now that TA0 and TA1 are unified.
- Remove get_pdev_same_bus(), since in get_dimm_config() the
variable "pvt->pci_ta" for KNL is also ready, we can simply use
pci_read_config_dword(pvt->pci_ta, KNL_MCMTR, &pvt->info.mcmtr) to read
MCMTR.
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: https://lkml.kernel.org/r/57884350.1030401@supermicro.com
Link: http://lkml.kernel.org/r/20170523000910.87925-1-qiuxu.zhuo@intel.com
[ Make __populate_dimms() return int. ]
Signed-off-by: Borislav Petkov <bp@suse.de>
We don't need this quirk anymore now that the EDAC memory controller
representation matches the hardware.
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20170523000834.87881-1-qiuxu.zhuo@intel.com
[ Commit message. ]
Signed-off-by: Borislav Petkov <bp@suse.de>
Tony pointed out: "currently the driver pretends there is one big
8-channel memory controller per socket instead of 2 4-channel
controllers. This is fine with all memory controller populated with
symmetrical DIMM configurations, but runs into difficulties on
asymmetrical setups".
Restructure the driver to assign an EDAC memory controller to each real
h/w memory controller to resolve the issue.
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20170523000731.87793-1-qiuxu.zhuo@intel.com
[ Break some lines at convenient points. ]
Signed-off-by: Borislav Petkov <bp@suse.de>
EDAC assigns logical memory controller numbers in the order that we find
memory controllers, which depends on which PCI bus they are on. Some
systems end up with MC0 on socket0, others (e.g Haswell) have MC0 on
socket3.
All this is made more confusing for users because we use the string
"Socket" while generating names for memory controllers, but the number
that we attach there is the memory controller number. E.g.
EDAC MC0: Giving out device to module sbridge_edac.c controller
Haswell Socket#0: DEV 0000:ff:12.0 (INTERRUPT)
Change the names to say "SrcID#%d" (where the number we use is read from
the h/w associated with the memory controller instead of some logical
number internal to the EDAC driver). New message:
EDAC MC0: Giving out device to module sbridge_edac.c controller
Haswell SrcID#3: DEV 0000:ff:12.0 (INTERRUPT)
Reported-by: Andrey Korolyov <andrey@xdel.ru>
Reported-by: Patrick Geary <patrickg@supermicro.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20170523000603.87748-1-qiuxu.zhuo@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
Each of the PCI device IDs belongs to a CPU socket, or to one of the
integrated memory controllers. Provide an enum to specify the domain of
each, and distinguish the resource number in each domain: the number
of the PCI device IDs per integrated memory controller/socket, and the
number of integrated memory controllers per socket.
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20170523000533.87704-1-qiuxu.zhuo@intel.com
[ Realign pci_dev_descr_knl members. ]
Signed-off-by: Borislav Petkov <bp@suse.de>
Pull RAS updates from Ingo Molnar:
"The main changes in this cycle were:
- Assign notifier chain priorities for all RAS related handlers to
make the ordering explicit (Borislav Petkov)
- Improve the AMD MCA banks sysfs output (Yazen Ghannam)
- Various cleanups and restructuring of the x86 RAS code (Borislav
Petkov)"
* 'ras-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/ras, EDAC, acpi: Assign MCE notifier handlers a priority
x86/ras: Get rid of mce_process_work()
EDAC/mce/amd: Dump TSC value
EDAC/mce/amd: Unexport amd_decode_mce()
x86/ras/amd/inj: Change dependency
x86/ras: Flip the TSC-adding logic
x86/ras/amd: Make sysfs names of banks more user-friendly
x86/ras/therm_throt: Do not log a fake MCE for thermal events
x86/ras/inject: Make it depend on X86_LOCAL_APIC=y
Assign all notifiers on the MCE decode chain a priority so that they get
called in the correct order.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Yazen Ghannam <Yazen.Ghannam@amd.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20170123183514.13356-10-bp@alien8.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Function sbridge_register_mci() sets pvt->info.show_interleave_mode
to knl_show_interleave_mode() on Knight's Landing and
show_interleave_mode() anywhere else.
Merge show_interleave_mode() and knl_show_interleave_mode() in a single
implementation and use it without an indirect function pointer.
Signed-off-by: Nicolas Iooss <nicolas.iooss_linux@m4x.org>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20170122172806.10412-1-nicolas.iooss_linux@m4x.org
[ Call it get_intlv_mode_str(). ]
Signed-off-by: Borislav Petkov <bp@suse.de>
Add Knights Mill (KNM) to the list of CPU models supported by sb_edac.
Signed-off-by: Piotr Luc <piotr.luc@intel.com>
Reviewed-by: Dave Hansen <dave.hansen@intel.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20161013153105.2517-6-piotr.luc@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
We now have symbolic names for a bunch of Intel CPU models via
asm/intel-family.h. The original conversion missed the EDAC drivers.
Convert them.
Signed-off-by: Dave Hansen <dave.hansen@intel.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20160929204321.9FAE5F84@viggo.jf.intel.com
[ Remove comment, macro name is descriptive enough. ]
Signed-off-by: Borislav Petkov <bp@suse.de>
buffers (Thor Thayer)
* Split the memory controller part out of mpc85xx and share it with a
* new Freescale ARM Layerscape driver (York Sun)
* amd64_edac fixes (Yazen Ghannam)
* Misc cleanups, refactoring and fixes all over the place
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=pcI8
-----END PGP SIGNATURE-----
Merge tag 'edac_for_4.9' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp
Pull EDAC updates from Borislav Petkov:
"A lot of movement in the EDAC tree this time around, coarse summary
below:
- Altera Arria10 enablement of NAND, DMA, USB, QSPI and SD-MMC FIFO
buffers (Thor Thayer)
- split the memory controller part out of mpc85xx and share it with a
new Freescale ARM Layerscape driver (York Sun)
- amd64_edac fixes (Yazen Ghannam)
- misc cleanups, refactoring and fixes all over the place"
* tag 'edac_for_4.9' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp: (37 commits)
EDAC, altera: Add IRQ Flags to disable IRQ while handling
EDAC, altera: Correct EDAC IRQ error message
EDAC, amd64: Autoload module using x86_cpu_id
EDAC, sb_edac: Remove NULL pointer check on array pci_tad
EDAC: Remove NO_IRQ from powerpc-only drivers
EDAC, fsl_ddr: Fix error return code in fsl_mc_err_probe()
EDAC, fsl_ddr: Add entry to MAINTAINERS
EDAC: Move Doug Thompson to CREDITS
EDAC, I3000: Orphan driver
EDAC, fsl_ddr: Replace simple_strtoul() with kstrtoul()
EDAC, layerscape: Add Layerscape EDAC support
EDAC, fsl_ddr: Fix IRQ dispose warning when module is removed
EDAC, fsl_ddr: Add support for little endian
EDAC, fsl_ddr: Add missing DDR DRAM types
EDAC, fsl_ddr: Rename macros and names
EDAC, fsl-ddr: Separate FSL DDR driver from MPC85xx
EDAC, mpc85xx: Replace printk() with pr_* format
EDAC, mpc85xx: Drop setting/clearing RFXE bit in HID1
EDAC, altera: Rename MC trigger to common name
EDAC, altera: Rename device trigger to common name
...
pvt->pci_tad is a NUM_CHANNELS array of struct pci_dev pointers and
hence cannot be NULL, so the NULL pointer check on pci_tad is redundant.
Remove it.
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Tony Luck <tony.luck@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/20160908083801.14766-1-colin.king@canonical.com
Signed-off-by: Borislav Petkov <bp@suse.de>
On Intel Xeon Phi Knights Landing processor family the channels of the
memory controller have untypical arrangement - MC0 is mapped to CH3,4,5
and MC1 is mapped to CH0,1,2. This causes the EDAC driver to report the
channel name incorrectly.
We missed this change earlier, so the code already contains similar
comment, but the translation function is incorrect.
Without this patch:
errors in DIMM_A and DIMM_D were reported in DIMM_D
errors in DIMM_B and DIMM_E were reported in DIMM_E
errors in DIMM_C and DIMM_F were reported in DIMM_F
Correct this.
Hubert Chrzaniuk:
- rebased to 4.8
- comments and code cleanup
Fixes: d0cdf90031 ("sb_edac: Add Knights Landing (Xeon Phi gen 2) support")
Reviewed-by: Tony Luck <tony.luck@intel.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Cc: lukasz.anaczkowski@intel.com
Cc: lukasz.odzioba@intel.com
Cc: mchehab@kernel.org
Cc: <stable@vger.kernel.org> # v4.5..
Link: http://lkml.kernel.org/r/1469231089-22837-1-git-send-email-lukasz.odzioba@intel.com
Signed-off-by: Lukasz Odzioba <lukasz.odzioba@intel.com>
[ Boris: Simplify a bit by removing char mc. ]
Signed-off-by: Borislav Petkov <bp@suse.de>
In commit 2c1ea4c700 ("EDAC, sb_edac: Use cpu family/model in driver
detection") I broke Knights Landing because I failed to notice that it
called a wrapper macro "sbridge_get_all_devices_knl" instead of
"sbridge_get_all_devices" like all the other types.
Now that we include the processor type in the pci_id_table structure we
can skip the wrappers and just have the sbridge_get_all_devices() check
the type to decide whether to allow duplicate devices and controllers to
have registers spread across buses.
Fixes: 2c1ea4c700 ("EDAC, sb_edac: Use cpu family/model in driver detection")
Tested-by: Lukasz Odzioba <lukasz.odzioba@intel.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In commit
2c1ea4c700 ("EDAC, sb_edac: Use cpu family/model in driver detection")
we switched from using PCI ids to determine which platform we are
running on to using CPU model instead.
I forgot that Broadwell-DE has its own distinct model number different
from Broadwell-EP or -EX.
Fixing this isn't just adding a line to the array of cpuids - the
exising code assumed a 1:1 mapping between entries in that array and the
"enum type" values. Added the type to pci_id_table structure to remove
this dependency and allows two Broadwell cpu models.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: Aristeu Rozanski <arozansk@redhat.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Fixes: 2c1ea4c700 ("EDAC, sb_edac: Use cpu family/model in driver detection")
Link: http://lkml.kernel.org/r/b3cffe40dec6dfe0235a5d52a504f0ba86a07ce7.1464902605.git.tony.luck@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
Broadwell made a small change to the rank target register moving the
target rank ID field up from bits 16:19 to bits 20:23.
Also found that the offset field grew by one bit in the IVY_BRIDGE to
HASWELL transition, so fix the RIR_OFFSET() macro too.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: stable@vger.kernel.org # v3.19+
Cc: Aristeu Rozanski <arozansk@redhat.com>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/2943fb819b1f7e396681165db9c12bb3df0e0b16.1464735623.git.tony.luck@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
* Remove ad-hoc buffering of MCE records in sb_edac and i7core_edac. (Tony Luck)
* Do not register sb_edac with pci_register_driver(). (Tony Luck)
* Add support for Skylake to ie31200_edac. (Jason Baron)
* Do not register amd64_edac with pci_register_driver(). (Borislav Petkov)
+ the usual round of cleanups and fixes all over the place.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=ud2P
-----END PGP SIGNATURE-----
Merge tag 'edac_for_4.7' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp
Pull EDAC updates from Borislav Petkov:
"It was pretty busy in EDAC land this time:
- Altera Arria10 L2 cache and On-Chip RAM ECC handling (Thor Thayer)
- Remove ad-hoc buffering of MCE records in sb_edac and i7core_edac
(Tony Luck)
- Do not register sb_edac with pci_register_driver() (Tony Luck)
- Add support for Skylake to ie31200_edac (Jason Baron)
- Do not register amd64_edac with pci_register_driver() (Borislav
Petkov)
... plus the usual round of cleanups and fixes all over the place"
* tag 'edac_for_4.7' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp: (25 commits)
EDAC, amd64_edac: Drop pci_register_driver() use
EDAC, ie31200_edac: Add Skylake support
EDAC, sb_edac: Use cpu family/model in driver detection
EDAC, i7core: Remove double buffering of error records
EDAC, amd64_edac: Issue driver banner only on success
ARM: socfpga: Initialize Arria10 OCRAM ECC on startup
EDAC: Increment correct counter in edac_inc_ue_error()
EDAC, sb_edac: Remove double buffering of error records
EDAC: Fix used after kfree() error in edac_unregister_sysfs()
EDAC, altera: Avoid unused function warnings
EDAC, altera: Remove useless casts
ARM: socfpga: Enable Arria10 OCRAM ECC on startup
EDAC, altera: Add Arria10 OCRAM ECC support
Documentation: dt: socfpga: Add Altera Arria10 OCRAM binding
EDAC, altera: Make OCRAM ECC dependency check generic
EDAC, altera: Add register offset for ECC Enable
EDAC, altera: Extract error inject operations to a struct fops
ARM: socfpga: Enable Arria10 L2 cache ECC on startup
EDAC, altera: Add Arria10 L2 Cache ECC handling
Documentation, dt, socfpga: Add Altera Arria10 L2 cache binding
...
Instead of picking a random PCI ID from the dozen or so we need to
access, just use x86_match_cpu() to pick based on CPU model number. The
choosing of PCI devices has been problematic in the past, see
11249e7399 ("sb_edac: Fix detection on SNB machines")
which fixed problems introduced by
d0585cd815 ("sb_edac: Claim a different PCI device").
This is especially ugly if future hardware might not even have
EDAC-relevant registers in PCI config space and we would still be
required to choose some "random" PCI devices to scan for just so our
driver loads.
Is this cleaner/clearer? It deletes much more code than it adds. Only
tested on Broadwell. The driver loads/unloads and loads again. Still
decodes errors too.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Suggested-by: Borislav Petkov <bp@alien8.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Both of these drivers can return NOTIFY_BAD, but this terminates
processing other callbacks that were registered later on the chain.
Since the driver did nothing to log the error it seems wrong to prevent
other interested parties from seeing it. E.g. neither of them had even
bothered to check the type of the error to see if it was a memory error
before the return NOTIFY_BAD.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Acked-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/72937355dd92318d2630979666063f8a2853495b.1461864507.git.tony.luck@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
In the bad old days the functions from x86_mce_decoder_chain could be
called in machine check context. So we used to carefully copy them and
defer processing until later. But in
f29a7aff4b ("x86/mce: Avoid potential deadlock due to printk() in MCE context")
we switched the logging code to save the record in a genpool, and call
the functions that registered to be notified later from a work queue.
So drop all the double buffering and do all the work we want to do as
soon as sbridge_mce_check_error() is called.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: Aristeu Rozanski <arozansk@redhat.com>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Cc: patrickg@supermicro.com
Link: http://lkml.kernel.org/r/100025611cd780d9bca72792b2b2146760da53e0.1460756761.git.tony.luck@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
Haswell and Broadwell can be configured to hash the channel
interleave function using bits [27:12] of the physical address.
On those processor models we must check to see if hashing is
enabled (bit21 of the HASWELL_HASYSDEFEATURE2 register) and
act accordingly.
Based on a patch by patrickg <patrickg@supermicro.com>
Tested-by: Patrick Geary <patrickg@supermicro.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: Aristeu Rozanski <arozansk@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-edac@vger.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
In commit:
eb1af3b71f ("Fix computation of channel address")
I switched the "sck_way" variable from holding the log2 value read
from the h/w to instead be the actual number. Unfortunately it
is needed in log2 form when used to shift the address.
Tested-by: Patrick Geary <patrickg@supermicro.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: Aristeu Rozanski <arozansk@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-edac@vger.kernel.org
Cc: stable@vger.kernel.org
Fixes: eb1af3b71f ("Fix computation of channel address")
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull RAS updates from Ingo Molnar:
"Various RAS updates:
- AMD MCE support updates for future CPUs, fixes and 'SMCA' (Scalable
MCA) error decoding support (Aravind Gopalakrishnan)
- x86 memcpy_mcsafe() support, to enable smart(er) hardware error
recovery in NVDIMM drivers, based on an extension of the x86
exception handling code. (Tony Luck)"
* 'ras-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
EDAC/sb_edac: Fix computation of channel address
x86/mm, x86/mce: Add memcpy_mcsafe()
x86/mce/AMD: Document some functionality
x86/mce: Clarify comments regarding deferred error
x86/mce/AMD: Fix logic to obtain block address
x86/mce/AMD, EDAC: Enable error decoding of Scalable MCA errors
x86/mce: Move MCx_CONFIG MSR definitions
x86/mce: Check for faults tagged in EXTABLE_CLASS_FAULT exception table entries
x86/mm: Expand the exception table logic to allow new handling options
x86/mce/AMD: Set MCAX Enable bit
x86/mce/AMD: Carve out threshold block preparation
x86/mce/AMD: Fix LVT offset configuration for thresholding
x86/mce/AMD: Reduce number of blocks scanned per bank
x86/mce/AMD: Do not perform shared bank check for future processors
x86/mce: Fix order of AMD MCE init function call
Large memory Haswell-EX systems with multiple DIMMs per channel were
sometimes reporting the wrong DIMM.
Found three problems:
1) Debug printouts for socket and channel interleave were not interpreting
the register fields correctly. The socket interleave field is a 2^X
value (0=1, 1=2, 2=4, 3=8). The channel interleave is X+1 (0=1, 1=2,
2=3. 3=4).
2) Actual use of the socket interleave value didn't interpret as 2^X
3) Conversion of address to channel address was complicated, and wrong.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: Aristeu Rozanski <arozansk@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-edac@vger.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Correct a typo introduced by
d0cdf90031 ("EDAC, sb_edac: Add Knights Landing (Xeon Phi gen 2) support")
As a result under some configurations DIMMs were not correctly
recognized. Problem affects only Xeon Phi architecture.
Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Cc: lukasz.anaczkowski@intel.com
Link: http://lkml.kernel.org/r/1457361045-26221-1-git-send-email-hubert.chrzaniuk@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
Knights Landing does not come with register that could be used to fetch
DIMM width. However the value is fixed for this architecture so it can
be hardcoded.
Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com>
Cc: Doug Thompson <dougthompson@xmission.com>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Cc: lukasz.anaczkowski@intel.com
Link: http://lkml.kernel.org/r/1449840082-18673-1-git-send-email-hubert.chrzaniuk@intel.com
Signed-off-by: Borislav Petkov <bp@suse.de>
Knights Landing is the next generation architecture for HPC market.
KNL introduces concept of a tile and CHA - Cache/Home Agent for memory
accesses.
Some things are fixed in KNL:
() There's single DIMM slot per channel
() There's 2 memory controllers with 3 channels each, however,
from EDAC standpoint, it is presented as single memory controller
with 6 channels. In order to represent 2 MCs w/ 3 CH, it would
require major redesign of EDAC core driver.
Basically, two functionalities are added/extended:
() during driver initialization KNL topology is being recognized, i.e.
which channels are populated with what DIMM sizes
(knl_get_dimm_capacity function)
() handle MCE errors - channel swizzling
Reviewed-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Jim Snow <jim.m.snow@intel.com>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Cc: lukasz.anaczkowski@intel.com
Link: http://lkml.kernel.org/r/1449136134-23706-5-git-send-email-hubert.chrzaniuk@intel.com
[ Rebase to 4.4-rc3. ]
Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Add options to sbridge_get_all_devices() to allow for duplicate device
IDs and devices that are scattered across mulitple PCI buses.
Signed-off-by: Jim Snow <jim.m.snow@intel.com>
Acked-by: Tony Luck <tony.luck@intel.com>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Cc: lukasz.anaczkowski@intel.com
Link: http://lkml.kernel.org/r/1449136134-23706-4-git-send-email-hubert.chrzaniuk@intel.com
[ Rebase to 4.4-rc3. ]
Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
SAD limit, interleave mode and DRAM related functionalities are now
virtualized, so that overriding them is easier.
Signed-off-by: Jim Snow <jim.m.snow@intel.com>
Acked-by: Tony Luck <tony.luck@intel.com>
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Cc: lukasz.anaczkowski@intel.com
Link: http://lkml.kernel.org/r/1449136134-23706-3-git-send-email-hubert.chrzaniuk@intel.com
[ Rebase to 4.4-rc3. ]
Signed-off-by: Hubert Chrzaniuk <hubert.chrzaniuk@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
In commit
7d375bffa5 ("sb_edac: Fix support for systems with two home agents per socket")
NUM_CHANNELS was changed to 8 and the channel space was renumerated to
handle EN, EP, and EX configurations.
The *_mci_bind_devs() functions - except for sbridge_mci_bind_devs() -
got a new device presence check in the form of saw_chan_mask. However,
sbridge_mci_bind_devs() still uses the NUM_CHANNELS for loop.
With the increase in NUM_CHANNELS, this loop fails at index 4 since
SB only has 4 TADs. This results in the following error on SB machines:
EDAC sbridge: Some needed devices are missing
EDAC sbridge: Couldn't find mci handler
EDAC sbridge: Couldn't find mci handle
This patch adapts the saw_chan_mask logic for sbridge_mci_bind_devs() as
well.
After this patch:
EDAC MC0: Giving out device to module sbridge_edac.c controller Sandy Bridge Socket#0: DEV 0000:3f:0e.0 (POLLED)
EDAC MC1: Giving out device to module sbridge_edac.c controller Sandy Bridge Socket#1: DEV 0000:7f:0e.0 (POLLED)
Signed-off-by: Seth Jennings <sjenning@redhat.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Acked-by: Tony Luck <tony.luck@intel.com>
Tested-by: Borislav Petkov <bp@suse.de>
Cc: <stable@vger.kernel.org> # v4.2
Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-edac <linux-edac@vger.kernel.org>
Link: http://lkml.kernel.org/r/1438798561-10180-1-git-send-email-sjenning@redhat.com
Signed-off-by: Borislav Petkov <bp@suse.de>
dimm_dev_type has been incorrectly determined in sb_edac. This patch fixes it
for Ivy Bridge and Haswell only since nothing like exists for Sandy Bridge.
We tested this patch in multiple systems matching the results with the
installed memory modules.
Acked-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
In case the memory banks are populated so the first channel isn't used, the
DDRIO PCI device won't be visible and it won't be possible to determine the
memory type.
Acked-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Basic support for the single socket Broadwell-DE processor
was added back in commit 1f39581a9a
sb_edac: Add support for Broadwell-DE processor
This patch extends Broadwell support to cover the two
socket "-EP" and four socket "-EX" versions of Broadwell.
Only tested on the 2 socket - but this code is largely
cloned from the Haswell path.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
First noticed a problem on a 4 socket machine where EDAC only reported
half the DIMMS. Tracked this down to the code that assumes that systems
with two home agents only have two memory channels on each agent. This
is true on 2 sockect ("-EP") machines. But four socket ("-EX") machines
have four memory channels on each home agent.
The old code would have had problems on two socket systems as it did
a shuffling trick to make the internals of the code think that the
channels from the first agent were '0' and '1', with the second agent
providing '2' and '3'. But the code didn't uniformly convert from
{ha,channel} tuples to this internal representation.
New code always considers up to eight channels.
On a machine with a single home agent these map easily to edac channels
0, 1, 2, 3. On machines with two home agents we map using:
edac_channel = 4*ha# + channel
So on a -EP machine where each home agent supports only two channels
we'll fill in channels 0, 1, 4, 5, and on a -EX machine we use all of 0,
1, 2, 3, 4, 5, 6, 7.
[mchehab@osg.samsung.com: fold a fixup patch as per Tony's request and fixed
a few CodingStyle issues]
Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
typo: "a7mode" chooses whether to use bits {8, 7, 9} or {8, 7, 6}
in the algorithm to spread access between memory resources. But
the non-a7mode path was incorrectly using GET_BITFIELD(addr, 7, 9)
and so picking bits {9, 8, 7}
thinko: BIT(1) of the dram_rule registers chooses whether to just
use the {8, 7, 6} (or {8, 7, 9}) bits mentioned above as they are,
or to XOR them with bits {18, 17, 16} but the code inverted the
test. We need the additional XOR when dram_rule{1} == 0.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
d0585cd815 ("sb_edac: Claim a different PCI device") changed the
probing of sb_edac to look for PCI device 0x3ca0:
3f:0e.0 System peripheral: Intel Corporation Xeon E5/Core i7 Processor Home Agent (rev 07)
00: 86 80 a0 3c 00 00 00 00 07 00 80 08 00 00 80 00
...
but we're matching for 0x3ca8, i.e. PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_TA
in sbridge_probe() therefore the probing fails.
Changing it to probe for 0x3ca0 (PCI_DEVICE_ID_INTEL_SBRIDGE_IMC_HA0),
.i.e., the 14.0 device, fixes the issue and driver loads successfully
again:
[ 2449.013120] EDAC DEBUG: sbridge_init:
[ 2449.017029] EDAC sbridge: Seeking for: PCI ID 8086:3ca0
[ 2449.022368] EDAC DEBUG: sbridge_get_onedevice: Detected 8086:3ca0
[ 2449.028498] EDAC sbridge: Seeking for: PCI ID 8086:3ca0
[ 2449.033768] EDAC sbridge: Seeking for: PCI ID 8086:3ca8
[ 2449.039028] EDAC DEBUG: sbridge_get_onedevice: Detected 8086:3ca8
[ 2449.045155] EDAC sbridge: Seeking for: PCI ID 8086:3ca8
...
Add a debug printk while at it to be able to catch the failure in the
future and dump driver version on successful load.
Fixes: d0585cd815 ("sb_edac: Claim a different PCI device")
Cc: stable@vger.kernel.org # 3.18
Acked-by: Aristeu Rozanski <aris@redhat.com>
Cc: Tony Luck <tony.luck@intel.com>
Acked-by: Andy Lutomirski <luto@amacapital.net>
Acked-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Code will always think there are 16 banks because of a typo
Reported-by: Misha
Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Broadwell-DE is the microserver version of next generation Xeon
processors. A whole bunch of new PCIe device ids, but otherwise
pretty much the same as Haswell.
Acked-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Haswell moved the TOLM/TOHM registers to a different device and offset.
The sb_edac driver accounted for the change of device, but not for the
new offset. There was also a typo in the constant to fill in the low
26 bits (was 0x1ffffff, should be 0x3ffffff).
This resulted in a bogus value for the top of low memory:
EDAC DEBUG: get_memory_layout: TOLM: 0.032 GB (0x0000000001ffffff)
which would result in EDAC refusing to translate addresses for
errors above the bogus value and below 4GB:
sbridge MC3: HANDLING MCE MEMORY ERROR
sbridge MC3: CPU 0: Machine Check Event: 0 Bank 7: 8c00004000010090
sbridge MC3: TSC 0
sbridge MC3: ADDR 2000000
sbridge MC3: MISC 523eac86
sbridge MC3: PROCESSOR 0:306f3 TIME 1414600951 SOCKET 0 APIC 0
MC3: 1 CE Error at TOLM area, on addr 0x02000000 on any memory ( page:0x0 offset:0x0 grain:32 syndrome:0x0)
With the fix we see the correct TOLM value:
DEBUG: get_memory_layout: TOLM: 2.048 GB (0x000000007fffffff)
and we decode address 2000000 correctly:
sbridge MC3: HANDLING MCE MEMORY ERROR
sbridge MC3: CPU 0: Machine Check Event: 0 Bank 7: 8c00004000010090
sbridge MC3: TSC 0
sbridge MC3: ADDR 2000000
sbridge MC3: MISC 523e1086
sbridge MC3: PROCESSOR 0:306f3 TIME 1414601319 SOCKET 0 APIC 0
DEBUG: get_memory_error_data: SAD interleave package: 0 = CPU socket 0, HA 0, shiftup: 0
DEBUG: get_memory_error_data: TAD#0: address 0x0000000002000000 < 0x000000007fffffff, socket interleave 1, channel interleave 4 (offset 0x00000000), index 0, base ch: 0, ch mask: 0x01
DEBUG: get_memory_error_data: RIR#0, limit: 4.095 GB (0x00000000ffffffff), way: 1
DEBUG: get_memory_error_data: RIR#0: channel address 0x00200000 < 0xffffffff, RIR interleave 0, index 0
DEBUG: sbridge_mce_output_error: area:DRAM err_code:0001:0090 socket:0 channel_mask:1 rank:0
MC3: 1 CE memory read error on CPU_SrcID#0_Channel#0_DIMM#0 (channel:0 slot:0 page:0x2000 offset:0x0 grain:32 syndrome:0x0 - area:DRAM err_code:0001:0090 socket:0 channel_mask:1 rank:0)
Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
sb_edac controls a large number of different PCI functions. Rather
than registering as a normal PCI driver for all of them, it
registers for just one so that it gets probed and, at probe time, it
looks for all the others.
Coincidentally, the device it registers for also contains the SMBUS
registers, so the PCI core will refuse to probe both sb_edac and a
future iMC SMBUS driver. The drivers don't actually conflict, so
just change sb_edac's device table to probe a different device.
An alternative fix would be to merge the two drivers, but sb_edac
will also refuse to load on non-ECC systems, whereas i2c_imc would
still be useful without ECC.
The only user-visible change should be that sb_edac appears to bind
a different device.
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Cc: Rui Wang <ruiv.wang@gmail.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
The i2c_imc driver will use two of them, and moving only part of
the list seems messier.
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Intel IA32 SDM Table 15-14 defines channel 0xf as 'not specified', but
EDAC doesn't know about this and returns and INTERNAL ERROR when the
channel is greater than NUM_CHANNELS:
kernel: [ 1538.886456] CPU 0: Machine Check Exception: 0 Bank 1: 940000000000009f
kernel: [ 1538.886669] TSC 2bc68b22e7e812 ADDR 46dae7000 MISC 0 PROCESSOR 0:306e4 TIME 1390414572 SOCKET 0 APIC 0
kernel: [ 1538.971948] EDAC MC1: INTERNAL ERROR: channel value is out of range (15 >= 4)
kernel: [ 1538.972203] EDAC MC1: 0 CE memory read error on unknown memory (slot:0 page:0x46dae7 offset:0x0 grain:0 syndrome:0x0 - area:DRAM err_code:0000:009f socket:1 channel_mask:1 rank:0)
This commit changes sb_edac to forward a channel of -1 to EDAC if the
channel is not specified. edac_mc_handle_error() sets the channel to -1
internally after the error message anyway, so this commit should have no
effect other than avoiding the INTERNAL ERROR message when the channel
is not specified.
Signed-off-by: Seth Jennings <sjenning@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Haswell memory controllers are very similar to Ivy Bridge and Sandy Bridge
ones. This patch adds support to Haswell based systems.
[m.chehab@samsung.com: Fix CodingStyle issues]
Cc: Tony Luck <tony.luck@intel.com>
Signed-off-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <m.chehab@samsung.com>
