commit 281326be67 upstream.
The current single-bit error injection mechanism flips bits directly in ECC RAM
by performing write and read operations. When the ECC RAM is actively used by
the Ethernet or USB controller, this approach sometimes trigger a false
double-bit error.
Switch both Ethernet and USB EDAC devices to use the INTTEST register
(altr_edac_a10_device_inject_fops) for single-bit error injection, similar to
the existing double-bit error injection method.
Fixes: 064acbd4f4 ("EDAC, altera: Add Stratix10 peripheral support")
Signed-off-by: Niravkumar L Rabara <niravkumarlaxmidas.rabara@altera.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Acked-by: Dinh Nguyen <dinguyen@kernel.org>
Cc: stable@vger.kernel.org
Link: https://patch.msgid.link/20251111081333.1279635-1-niravkumarlaxmidas.rabara@altera.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 6e1c2c6c2c ]
Newer AMD systems can support up to 16 channels per EDAC "mc" device.
These are detected by the EDAC module running on the device, and the
current EDAC interface is appropriately enumerated.
The legacy EDAC sysfs interface however, provides device attributes for
channels 0 through 11 only. Consequently, the last four channels, 12
through 15, will not be enumerated and will not be visible through the
legacy sysfs interface.
Add additional device attributes to ensure that all 16 channels, if
present, are enumerated by and visible through the legacy EDAC sysfs
interface.
Signed-off-by: Avadhut Naik <avadhut.naik@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/20250916203242.1281036-1-avadhut.naik@amd.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 2e6fe1bbef ]
When loading the i10nm_edac driver on some Intel Granite Rapids servers,
a call trace may appear as follows:
UBSAN: shift-out-of-bounds in drivers/edac/skx_common.c:453:16
shift exponent -66 is negative
...
__ubsan_handle_shift_out_of_bounds+0x1e3/0x390
skx_get_dimm_info.cold+0x47/0xd40 [skx_edac_common]
i10nm_get_dimm_config+0x23e/0x390 [i10nm_edac]
skx_register_mci+0x159/0x220 [skx_edac_common]
i10nm_init+0xcb0/0x1ff0 [i10nm_edac]
...
This occurs because some BIOS may disable a memory controller if there
aren't any memory DIMMs populated on this memory controller. The DIMMMTR
register of this disabled memory controller contains the invalid value
~0, resulting in the call trace above.
Fix this call trace by skipping DIMM enumeration on a disabled memory
controller.
Fixes: ba987eaaab ("EDAC/i10nm: Add Intel Granite Rapids server support")
Reported-by: Jose Jesus Ambriz Meza <jose.jesus.ambriz.meza@intel.com>
Reported-by: Chia-Lin Kao (AceLan) <acelan.kao@canonical.com>
Closes: https://lore.kernel.org/all/20250730063155.2612379-1-acelan.kao@canonical.com/
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Tested-by: Chia-Lin Kao (AceLan) <acelan.kao@canonical.com>
Link: https://lore.kernel.org/r/20250806065707.3533345-1-qiuxu.zhuo@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1a251f52cf ]
This just standardizes the use of MIN() and MAX() macros, with the very
traditional semantics. The goal is to use these for C constant
expressions and for top-level / static initializers, and so be able to
simplify the min()/max() macros.
These macro names were used by various kernel code - they are very
traditional, after all - and all such users have been fixed up, with a
few different approaches:
- trivial duplicated macro definitions have been removed
Note that 'trivial' here means that it's obviously kernel code that
already included all the major kernel headers, and thus gets the new
generic MIN/MAX macros automatically.
- non-trivial duplicated macro definitions are guarded with #ifndef
This is the "yes, they define their own versions, but no, the include
situation is not entirely obvious, and maybe they don't get the
generic version automatically" case.
- strange use case #1
A couple of drivers decided that the way they want to describe their
versioning is with
#define MAJ 1
#define MIN 2
#define DRV_VERSION __stringify(MAJ) "." __stringify(MIN)
which adds zero value and I just did my Alexander the Great
impersonation, and rewrote that pointless Gordian knot as
#define DRV_VERSION "1.2"
instead.
- strange use case #2
A couple of drivers thought that it's a good idea to have a random
'MIN' or 'MAX' define for a value or index into a table, rather than
the traditional macro that takes arguments.
These values were re-written as C enum's instead. The new
function-line macros only expand when followed by an open
parenthesis, and thus don't clash with enum use.
Happily, there weren't really all that many of these cases, and a lot of
users already had the pattern of using '#ifndef' guarding (or in one
case just using '#undef MIN') before defining their own private version
that does the same thing. I left such cases alone.
Cc: David Laight <David.Laight@aculab.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Eliav Farber <farbere@amazon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 4477b39c32 ]
Commit 3a7e02c040 ("minmax: avoid overly complicated constant
expressions in VM code") added the simpler MIN_T/MAX_T macros in order
to avoid some excessive expansion from the rather complicated regular
min/max macros.
The complexity of those macros stems from two issues:
(a) trying to use them in situations that require a C constant
expression (in static initializers and for array sizes)
(b) the type sanity checking
and MIN_T/MAX_T avoids both of these issues.
Now, in the whole (long) discussion about all this, it was pointed out
that the whole type sanity checking is entirely unnecessary for
min_t/max_t which get a fixed type that the comparison is done in.
But that still leaves min_t/max_t unnecessarily complicated due to
worries about the C constant expression case.
However, it turns out that there really aren't very many cases that use
min_t/max_t for this, and we can just force-convert those.
This does exactly that.
Which in turn will then allow for much simpler implementations of
min_t()/max_t(). All the usual "macros in all upper case will evaluate
the arguments multiple times" rules apply.
We should do all the same things for the regular min/max() vs MIN/MAX()
cases, but that has the added complexity of various drivers defining
their own local versions of MIN/MAX, so that needs another level of
fixes first.
Link: https://lore.kernel.org/all/b47fad1d0cf8449886ad148f8c013dae@AcuMS.aculab.com/
Cc: David Laight <David.Laight@aculab.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Eliav Farber <farbere@amazon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit a3f3040657 upstream.
Each Chip-Select (CS) of a Unified Memory Controller (UMC) on AMD Zen-based
SOCs has an Address Mask and a Secondary Address Mask register associated with
it. The amd64_edac module logs DIMM sizes on a per-UMC per-CS granularity
during init using these two registers.
Currently, the module primarily considers only the Address Mask register for
computing DIMM sizes. The Secondary Address Mask register is only considered
for odd CS. Additionally, if it has been considered, the Address Mask register
is ignored altogether for that CS. For power-of-two DIMMs i.e. DIMMs whose
total capacity is a power of two (32GB, 64GB, etc), this is not an issue
since only the Address Mask register is used.
For non-power-of-two DIMMs i.e., DIMMs whose total capacity is not a power of
two (48GB, 96GB, etc), however, the Secondary Address Mask register is used
in conjunction with the Address Mask register. However, since the module only
considers either of the two registers for a CS, the size computed by the
module is incorrect. The Secondary Address Mask register is not considered for
even CS, and the Address Mask register is not considered for odd CS.
Introduce a new helper function so that both Address Mask and Secondary
Address Mask registers are considered, when valid, for computing DIMM sizes.
Furthermore, also rename some variables for greater clarity.
Fixes: 81f5090db8 ("EDAC/amd64: Support asymmetric dual-rank DIMMs")
Closes: https://lore.kernel.org/dbec22b6-00f2-498b-b70d-ab6f8a5ec87e@natrix.lt
Reported-by: Žilvinas Žaltiena <zilvinas@natrix.lt>
Signed-off-by: Avadhut Naik <avadhut.naik@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Yazen Ghannam <yazen.ghannam@amd.com>
Tested-by: Žilvinas Žaltiena <zilvinas@natrix.lt>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/20250529205013.403450-1-avadhut.naik@amd.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit eeed3e03f4 ]
When enabling the retry_rd_err_log (RRL) feature during the loading of the
i10nm_edac driver with the module parameter retry_rd_err_log=2 (Linux RRL
control mode), the default values of the control bits of RRL are saved so
that they can be restored during the unloading of the driver.
In the current code, the RRL of pseudo channel 1 of HBM overwrites pseudo
channel 0 during the loading of the driver, resulting in the loss of saved
RRL for pseudo channel 0. This causes the RRL of pseudo channel 0 of HBM to
be wrongly restored with the values from pseudo channel 1 when unloading
the driver.
Fix this issue by creating two separate groups of RRL control registers
per channel to save default RRL settings of two {sub-,pseudo-}channels.
Fixes: acd4cf68fe ("EDAC/i10nm: Retrieve and print retry_rd_err_log registers for HBM")
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Tested-by: Feng Xu <feng.f.xu@intel.com>
Link: https://lore.kernel.org/r/20250417150724.1170168-3-qiuxu.zhuo@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 20d2d476b3 ]
After loading i10nm_edac (which automatically loads skx_edac_common), if
unload only i10nm_edac, then reload it and perform error injection testing,
a general protection fault may occur:
mce: [Hardware Error]: Machine check events logged
Oops: general protection fault ...
...
Workqueue: events mce_gen_pool_process
RIP: 0010:string+0x53/0xe0
...
Call Trace:
<TASK>
? die_addr+0x37/0x90
? exc_general_protection+0x1e7/0x3f0
? asm_exc_general_protection+0x26/0x30
? string+0x53/0xe0
vsnprintf+0x23e/0x4c0
snprintf+0x4d/0x70
skx_adxl_decode+0x16a/0x330 [skx_edac_common]
skx_mce_check_error.part.0+0xf8/0x220 [skx_edac_common]
skx_mce_check_error+0x17/0x20 [skx_edac_common]
...
The issue arose was because the variable 'adxl_component_count' (inside
skx_edac_common), which counts the ADXL components, was not reset. During
the reloading of i10nm_edac, the count was incremented by the actual number
of ADXL components again, resulting in a count that was double the real
number of ADXL components. This led to an out-of-bounds reference to the
ADXL component array, causing the general protection fault above.
Fix this issue by resetting the 'adxl_component_count' in adxl_put(),
which is called during the unloading of {skx,i10nm}_edac.
Fixes: 123b158635 ("EDAC, i10nm: make skx_common.o a separate module")
Reported-by: Feng Xu <feng.f.xu@intel.com>
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Tested-by: Feng Xu <feng.f.xu@intel.com>
Link: https://lore.kernel.org/r/20250417150724.1170168-2-qiuxu.zhuo@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit c29dfd661f ]
gcc-14 produces a bogus warning in some configurations:
drivers/edac/ie31200_edac.c: In function 'ie31200_probe1.isra':
drivers/edac/ie31200_edac.c:412:26: error: 'dimm_info' is used uninitialized [-Werror=uninitialized]
412 | struct dimm_data dimm_info[IE31200_CHANNELS][IE31200_DIMMS_PER_CHANNEL];
| ^~~~~~~~~
drivers/edac/ie31200_edac.c:412:26: note: 'dimm_info' declared here
412 | struct dimm_data dimm_info[IE31200_CHANNELS][IE31200_DIMMS_PER_CHANNEL];
| ^~~~~~~~~
I don't see any way the unintialized access could really happen here,
but I can see why the compiler gets confused by the two loops.
Instead, rework the two nested loops to only read the addr_decode
registers and then keep only one instance of the dimm info structure.
[Tony: Qiuxu pointed out that the "populate DIMM info" comment was left
behind in the refactor and suggested moving it. I deleted the comment
as unnecessry in front os a call to populate_dimm_info(). That seems
pretty self-describing.]
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Jason Baron <jbaron@akamai.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Link: https://lore.kernel.org/all/20250122065031.1321015-1-arnd@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit d9207cf776 ]
When doing error injection to some memory DIMMs on certain Intel Emerald
Rapids servers, the i10nm_edac missed error reports for some memory DIMMs.
Certain BIOS configurations may hide some memory controllers, and the
i10nm_edac doesn't enumerate these hidden memory controllers. However, the
ADXL decodes memory errors using memory controller physical indices even
if there are hidden memory controllers. Therefore, the memory controller
physical indices reported by the ADXL may mismatch the logical indices
enumerated by the i10nm_edac, resulting in missed error reports for some
memory DIMMs.
Fix this issue by creating a mapping table from memory controller physical
indices (used by the ADXL) to logical indices (used by the i10nm_edac) and
using it to convert the physical indices to the logical indices during the
error handling process.
Fixes: c545f5e412 ("EDAC/i10nm: Skip the absent memory controllers")
Reported-by: Kevin Chang <kevin1.chang@intel.com>
Tested-by: Kevin Chang <kevin1.chang@intel.com>
Reported-by: Thomas Chen <Thomas.Chen@intel.com>
Tested-by: Thomas Chen <Thomas.Chen@intel.com>
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Link: https://lore.kernel.org/r/20250214002728.6287-1-qiuxu.zhuo@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit c158647c10 upstream.
The previous implementation incorrectly configured the cmn_interrupt_2_enable
register for interrupt handling. Using cmn_interrupt_2_enable to configure
Tag, Data RAM ECC interrupts would lead to issues like double handling of the
interrupts (EL1 and EL3) as cmn_interrupt_2_enable is meant to be configured
for interrupts which needs to be handled by EL3.
EL1 LLCC EDAC driver needs to use cmn_interrupt_0_enable register to configure
Tag, Data RAM ECC interrupts instead of cmn_interrupt_2_enable.
Fixes: 27450653f1 ("drivers: edac: Add EDAC driver support for QCOM SoCs")
Signed-off-by: Komal Bajaj <quic_kbajaj@quicinc.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Cc: <stable@kernel.org>
Link: https://lore.kernel.org/r/20241119064608.12326-1-quic_kbajaj@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit fefaae9039 ]
The segmentation fault happens because:
During modprobe:
1. In igen6_probe(), igen6_pvt will be allocated with kzalloc()
2. In igen6_register_mci(), mci->pvt_info will point to
&igen6_pvt->imc[mc]
During rmmod:
1. In mci_release() in edac_mc.c, it will kfree(mci->pvt_info)
2. In igen6_remove(), it will kfree(igen6_pvt);
Fix this issue by setting mci->pvt_info to NULL to avoid the double
kfree.
Fixes: 10590a9d4f ("EDAC/igen6: Add EDAC driver for Intel client SoCs using IBECC")
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219360
Signed-off-by: Orange Kao <orange@aiven.io>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Link: https://lore.kernel.org/r/20241104124237.124109-2-orange@aiven.io
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit a36667037a ]
The Granite Rapids CPUs with Flat2LM memory configurations may
mistakenly report near-memory errors as far-memory errors, resulting
in the invalid decoded ADXL results:
EDAC skx: Bad imc -1
Fix this incorrect far-memory error source indicator by prefetching the
decoded far-memory controller ID, and adjust the error source indicator
to near-memory if the far-memory controller ID is invalid.
Fixes: ba987eaaab ("EDAC/i10nm: Add Intel Granite Rapids server support")
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Tested-by: Diego Garcia Rodriguez <diego.garcia.rodriguez@intel.com>
Link: https://lore.kernel.org/r/20241015072236.24543-3-qiuxu.zhuo@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 2397f79573 ]
The current skx_common determines whether the memory error source is the
near memory of the 2LM system and then retrieves the decoded error results
from the ADXL components (near-memory vs. far-memory) accordingly.
However, some memory controllers may have limitations in correctly
reporting the memory error source, leading to the retrieval of incorrect
decoded parts from the ADXL.
To address these limitations, instead of simply determining whether the
memory error is from the near memory of the 2LM system, it is necessary to
distinguish the memory error source details as follows:
Memory error from the near memory of the 2LM system.
Memory error from the far memory of the 2LM system.
Memory error from the 1LM system.
Not a memory error.
This will enable the i10nm_edac driver to take appropriate actions for
those memory controllers that have limitations in reporting the memory
error source.
Fixes: ba987eaaab ("EDAC/i10nm: Add Intel Granite Rapids server support")
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Tested-by: Diego Garcia Rodriguez <diego.garcia.rodriguez@intel.com>
Link: https://lore.kernel.org/r/20241015072236.24543-2-qiuxu.zhuo@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 9ec22ac4fe ]
Fix undefined behavior caused by left-shifting a negative value in the
expression:
cap_high ^ (1 << (bad_data_bit - 32))
The variable bad_data_bit ranges from 0 to 63. When it is less than 32,
bad_data_bit - 32 becomes negative, and left-shifting by a negative
value in C is undefined behavior.
Fix this by combining cap_high and cap_low into a 64-bit variable.
[ bp: Massage commit message, simplify error bits handling. ]
Fixes: ea2eb9a8b6 ("EDAC, fsl-ddr: Separate FSL DDR driver from MPC85xx")
Signed-off-by: Priyanka Singh <priyanka.singh@nxp.com>
Signed-off-by: Li Yang <leoyang.li@nxp.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20241016-imx95_edac-v3-3-86ae6fc2756a@nxp.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1fe774a93b ]
The 64-bit argument for the "get DIMM info" SMC call consists of mem_ctrl_idx
left-shifted 16 bits and OR-ed with DIMM index. With mem_ctrl_idx defined as
32-bits wide the left-shift operation truncates the upper 16 bits of
information during the calculation of the SMC argument.
The mem_ctrl_idx stack variable must be defined as 64-bits wide to prevent any
potential integer overflow, i.e. loss of data from upper 16 bits.
Fixes: 82413e562e ("EDAC, mellanox: Add ECC support for BlueField DDR4")
Signed-off-by: David Thompson <davthompson@nvidia.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Shravan Kumar Ramani <shravankr@nvidia.com>
Link: https://lore.kernel.org/r/20240930151056.10158-1-davthompson@nvidia.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 35e6dbfe18 ]
The Zynq UltraScale+ MPSoC DDR has a disjoint memory from 2GB to 32GB.
The DDR host interface has a contiguous memory so while injecting
errors, the driver should remove the hole else the injection fails as
the address translation is incorrect.
Introduce a get_mem_info() function pointer and set it for Zynq
UltraScale+ platform to return host address.
Fixes: 1a81361f75 ("EDAC, synopsys: Add Error Injection support for ZynqMP DDR controller")
Signed-off-by: Shubhrajyoti Datta <shubhrajyoti.datta@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240711100656.31376-1-shubhrajyoti.datta@amd.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 591c946675 ]
The race condition around the ECCCLR register access happens in the IRQ
disable method called in the device remove() procedure and in the ECC IRQ
handler:
1. Enable IRQ:
a. ECCCLR = EN_CE | EN_UE
2. Disable IRQ:
a. ECCCLR = 0
3. IRQ handler:
a. ECCCLR = CLR_CE | CLR_CE_CNT | CLR_CE | CLR_CE_CNT
b. ECCCLR = 0
c. ECCCLR = EN_CE | EN_UE
So if the IRQ disabling procedure is called concurrently with the IRQ
handler method the IRQ might be actually left enabled due to the
statement 3c.
The root cause of the problem is that ECCCLR register (which since
v3.10a has been called as ECCCTL) has intermixed ECC status data clear
flags and the IRQ enable/disable flags. Thus the IRQ disabling (clear EN
flags) and handling (write 1 to clear ECC status data) procedures must
be serialised around the ECCCTL register modification to prevent the
race.
So fix the problem described above by adding the spin-lock around the
ECCCLR modifications and preventing the IRQ-handler from modifying the
IRQs enable flags (there is no point in disabling the IRQ and then
re-enabling it again within a single IRQ handler call, see the
statements 3a/3b and 3c above).
Fixes: f7824ded41 ("EDAC/synopsys: Add support for version 3 of the Synopsys EDAC DDR")
Signed-off-by: Serge Semin <fancer.lancer@gmail.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240222181324.28242-2-fancer.lancer@gmail.com
Stable-dep-of: 35e6dbfe18 ("EDAC/synopsys: Fix error injection on Zynq UltraScale+")
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit e0d3350778 ]
There are no "struct page" associations with SGX pages, causing the check
pfn_to_online_page() to fail. This results in the inability to decode the
SGX addresses and warning messages like:
Invalid address 0x34cc9a98840 in IA32_MC17_ADDR
Add an additional check to allow the decoding of the error address and to
skip the warning message, if the error address is an SGX address.
Fixes: 1e92af09fa ("EDAC/skx_common: Filter out the invalid address")
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Link: https://lore.kernel.org/r/20240408120419.50234-1-qiuxu.zhuo@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 123b158635 ]
Commit 598afa0504 ("kbuild: warn objects shared among multiple modules")
was added to track down cases where the same object is linked into
multiple modules. This can cause serious problems if some modules are
builtin while others are not.
That test triggers this warning:
scripts/Makefile.build:236: drivers/edac/Makefile: skx_common.o is added to multiple modules: i10nm_edac skx_edac
Make this a separate module instead.
[Tony: Added more background details to commit message]
Fixes: d4dc89d069 ("EDAC, i10nm: Add a driver for Intel 10nm server processors")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Link: https://lore.kernel.org/all/20240529095132.1929397-1-arnd@kernel.org/
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit f8367a74ae upstream.
errcmd_enable_error_reporting() uses pci_{read,write}_config_word()
that return PCIBIOS_* codes. The return code is then returned all the
way into the probe function igen6_probe() that returns it as is. The
probe functions, however, should return normal errnos.
Convert PCIBIOS_* returns code using pcibios_err_to_errno() into normal
errno before returning it from errcmd_enable_error_reporting().
Fixes: 10590a9d4f ("EDAC/igen6: Add EDAC driver for Intel client SoCs using IBECC")
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240527132236.13875-2-ilpo.jarvinen@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3ec8ebd8a5 upstream.
gpu_get_node_map() uses pci_read_config_dword() that returns PCIBIOS_*
codes. The return code is then returned all the way into the module
init function amd64_edac_init() that returns it as is. The module init
functions, however, should return normal errnos.
Convert PCIBIOS_* returns code using pcibios_err_to_errno() into normal
errno before returning it from gpu_get_node_map().
For consistency, convert also the other similar cases which return
PCIBIOS_* codes even if they do not have any bugs at the moment.
Fixes: 4251566ebc ("EDAC/amd64: Cache and use GPU node map")
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240527132236.13875-1-ilpo.jarvinen@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 475c58e1a4 ]
Enabling -Wstringop-overflow globally exposes a warning for a common bug
in the usage of strncat():
drivers/edac/thunderx_edac.c: In function 'thunderx_ocx_com_threaded_isr':
drivers/edac/thunderx_edac.c:1136:17: error: 'strncat' specified bound 1024 equals destination size [-Werror=stringop-overflow=]
1136 | strncat(msg, other, OCX_MESSAGE_SIZE);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
...
1145 | strncat(msg, other, OCX_MESSAGE_SIZE);
...
1150 | strncat(msg, other, OCX_MESSAGE_SIZE);
...
Apparently the author of this driver expected strncat() to behave the
way that strlcat() does, which uses the size of the destination buffer
as its third argument rather than the length of the source buffer. The
result is that there is no check on the size of the allocated buffer.
Change it to strlcat().
[ bp: Trim compiler output, fixup commit message. ]
Fixes: 41003396f9 ("EDAC, thunderx: Add Cavium ThunderX EDAC driver")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://lore.kernel.org/r/20231122222007.3199885-1-arnd@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
Pull intel EDAC fixes from Tony Luck:
- Old igen6 driver could lose pending events during initialization
- Sapphire Rapids workstations have fewer memory controllers than their
bigger siblings. This confused the driver.
* tag 'edac_updates_for_v6.6' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras:
EDAC/igen6: Fix the issue of no error events
EDAC/i10nm: Skip the absent memory controllers
Pull devicetree include cleanups from Rob Herring:
"These are the remaining few clean-ups of DT related includes which
didn't get applied to subsystem trees"
* tag 'devicetree-header-cleanups-for-6.6' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
ipmi: Explicitly include correct DT includes
tpm: Explicitly include correct DT includes
lib/genalloc: Explicitly include correct DT includes
parport: Explicitly include correct DT includes
sbus: Explicitly include correct DT includes
mux: Explicitly include correct DT includes
macintosh: Explicitly include correct DT includes
hte: Explicitly include correct DT includes
EDAC: Explicitly include correct DT includes
clocksource: Explicitly include correct DT includes
sparc: Explicitly include correct DT includes
riscv: Explicitly include correct DT includes
The DT of_device.h and of_platform.h date back to the separate
of_platform_bus_type before it was merged into the regular platform bus.
As part of that merge prepping Arm DT support 13 years ago, they
"temporarily" include each other. They also include platform_device.h
and of.h. As a result, there's a pretty much random mix of those include
files used throughout the tree. In order to detangle these headers and
replace the implicit includes with struct declarations, users need to
explicitly include the correct includes.
Link: https://lore.kernel.org/r/20230714174434.4054728-1-robh@kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
Current igen6_edac checks for pending errors before the registration
of the error handler. However, there is a possibility that the error
occurs during the registration process, leading to unhandled pending
errors and no future error events. This issue can be reproduced by
repeatedly injecting errors during the loading of the igen6_edac.
Fix this issue by moving the pending error handler after the registration
of the error handler, ensuring that no pending errors are left unhandled.
Fixes: 10590a9d4f ("EDAC/igen6: Add EDAC driver for Intel client SoCs using IBECC")
Reported-by: Ee Wey Lim <ee.wey.lim@intel.com>
Tested-by: Ee Wey Lim <ee.wey.lim@intel.com>
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Link: https://lore.kernel.org/r/20230725080427.23883-1-qiuxu.zhuo@intel.com
Pull RAS updates from Borislav Petkov:
- Add initial support for RAS hardware found on AMD server GPUs (MI200).
Those GPUs and CPUs are connected together through the coherent
fabric and the GPU memory controllers report errors through x86's MCA
so EDAC needs to support them. The amd64_edac driver supports now HBM
(High Bandwidth Memory) and thus such heterogeneous memory controller
systems
- Other small cleanups and improvements
* tag 'ras_core_for_v6.5' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
EDAC/amd64: Cache and use GPU node map
EDAC/amd64: Add support for AMD heterogeneous Family 19h Model 30h-3Fh
EDAC/amd64: Document heterogeneous system enumeration
x86/MCE/AMD, EDAC/mce_amd: Decode UMC_V2 ECC errors
x86/amd_nb: Re-sort and re-indent PCI defines
x86/amd_nb: Add MI200 PCI IDs
ras/debugfs: Fix error checking for debugfs_create_dir()
x86/MCE: Check a hw error's address to determine proper recovery action
Pull EDAC updates from Borislav Petkov:
- amd64_edac: Add support for Zen4 client hardware
- amd64_edac: Remove the version string as it is useless and actively
confusing when looking at backported versions of the driver
- Add a driver for the Nuvoton NPCM memory controller
- A debugfs error checking cleanup
* tag 'edac_updates_for_v6.5' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras:
EDAC/npcm: Add NPCM memory controller driver
dt-bindings: memory-controllers: nuvoton: Add NPCM memory controller
EDAC/thunderx: Check debugfs file creation retval properly
EDAC/amd64: Add support for ECC on family 19h model 60h-7Fh
EDAC/amd64: Remove module version string
AMD systems have historically provided an "AMD Node ID" that is a unique
identifier for each die in a multi-die package. This was associated with
a unique instance of the AMD Northbridge on a legacy system. And now it
is associated with a unique instance of the AMD Data Fabric on modern
systems. Each instance is referred to as a "Node"; this is an
AMD-specific term not to be confused with NUMA nodes.
The data fabric provides a number of interfaces accessible through a set
of functions in a single PCI device. There is one PCI device per Data
Fabric (AMD Node), and multi-die systems will see multiple such PCI
devices. The AMD Node ID matches a Node's position in the PCI hierarchy.
For example, the Node 0 is accessed using the first PCI device, Node 1
is accessed using the second, and so on. A logical CPU can find its AMD
Node ID using CPUID. Furthermore, the AMD Node ID is used within the
hardware fabric, so it is not purely a logical value.
Heterogeneous AMD systems, with a CPU Data Fabric connected to GPU data
fabrics, follow a similar convention. Each CPU and GPU die has a unique
AMD Node ID value, and each Node ID corresponds to PCI devices in
sequential order.
However, there are two caveats:
1) GPUs are not x86, and they don't have CPUID to read their AMD Node ID
like on CPUs. This means the value is more implicit and based on PCI
enumeration and hardware-specifics.
2) There is a gap in the hardware values for AMD Node IDs. Values 0-7
are for CPUs and values 8-15 are for GPUs.
For example, a system with one CPU die and two GPUs dies will have the
following values:
CPU0 -> AMD Node 0
GPU0 -> AMD Node 8
GPU1 -> AMD Node 9
EDAC is the only subsystem where this has a practical effect. Memory
errors on AMD systems are commonly reported through MCA to a CPU on the
local AMD Node. The error information is passed along to EDAC where the
AMD EDAC modules use the AMD Node ID of reporting logical CPU to access
AMD Node information.
However, memory errors from a GPU die will be reported to the CPU die.
Therefore, the logical CPU's AMD Node ID can't be used since it won't
match the AMD Node ID of the GPU die. The AMD Node ID of the GPU die is
provided as part of the MCA information, and the value will match the
hardware enumeration (e.g. 8-15).
Handle this situation by discovering GPU dies the same way as CPU dies
in the AMD NB code. But do a "node id" fixup in AMD64 EDAC where it's
needed.
The GPU data fabrics provide a register with the base AMD Node ID for
their local "type", i.e. GPU data fabric. This value is the same for all
fabrics of the same type in a system.
Read and cache the base AMD Node ID from one of the GPU devices during
module initialization. Use this to fixup the "node id" when reporting
memory errors at runtime.
[ bp: Squash a fix making gpu_node_map static as reported by
Tom Rix <trix@redhat.com>.
Link: https://lore.kernel.org/r/20230610210930.174074-1-trix@redhat.com ]
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Co-developed-by: Muralidhara M K <muralidhara.mk@amd.com>
Signed-off-by: Muralidhara M K <muralidhara.mk@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230515113537.1052146-6-muralimk@amd.com
