It is helpful to see the saved record entries during run time in
human-readable format. This is useful for testing during module
development. It can also be used by system admins to quickly and easily
see the state of the system.
Provide a sequential file in debugfs to print fields of interest from
the FRU records and their entries.
Don't fail to load the module if the debugfs interface is not available.
This is a convenience feature which does not affect other module
functionality.
The new interface reads the record entries and should hold the mutex.
Expand the mutex code comment to clarify when it should be held.
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240301143748.854090-4-yazen.ghannam@amd.com
The system physical address (SPA) of an error is not a stable value. It
will change depending on the location of the memory: parts can be
swapped. And it will change depending on memory topology: NUMA nodes
and/or interleaving can be adjusted.
Therefore, the SPA value is not part of the "FRU Memory Poison" record
format. And it will not be saved to persistent storage.
However, the SPA values can be helpful during debug and for system
admins during run time.
Save the SPA values in a separate structure. This is updated when
records are restored and when new errors are saved.
[ bp: Make error messages more user friendly and add and correct
comments. ]
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240301143748.854090-3-yazen.ghannam@amd.com
The hash_pa8 and hashed_bit values in denorm_addr_df4_np2() are
currently defined as u8 types. These variables represent single bits.
'hash_pa8' is set based on logical AND operations using masks with more
than 8 bits. So the calculated value will not fit in this variable. It
will always be '0'. The 'hash_pa8' check later in the function will fail
which produces incorrect results for some cases.
Change these variables to bool type. This clarifies that they are
single bit values. Also, this allows the compiler to ensure they hold
the proper results. Remove an unnecessary shift operation.
[ bp: Remove the unnecessary brackets in the else-branch of the
hash_pa8 assignment. ]
Fixes: 3f3174996b ("RAS: Introduce AMD Address Translation Library")
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240222165449.23582-1-yazen.ghannam@amd.com
Memory errors are an expected occurrence on systems with high memory
density. Generally, errors within a small number of unique physical
locations are acceptable, based on manufacturer and/or admin policy.
During run time, memory with errors may be retired so it is no longer
used by the system. This is done in mm through page poisoning, and the
effect will remain until the system is restarted.
If a memory location is consistently faulty, then the same run time
error handling may occur in the next reboot cycle, leading to
terminating jobs due to that already known bad memory. This could be
prevented if information from the previous boot was not lost.
Some add-in cards with driver-managed memory have on-board persistent
storage. Their driver saves memory error information to the persistent
storage during run time. The information is then restored after reset,
and known bad memory will be retired before the hardware is used.
A running log of bad memory locations is kept across multiple resets.
A similar solution is desirable for CPUs. However, this solution should
leverage industry-standard components as much as possible, rather than
a bespoke platform driver.
Two components are needed: a record format and a persistent storage
interface.
Implement a new module to manage the record formats on persistent
storage. Use the requirements for an AMD MI300-based system to start.
Vendor- and platform-specific details can be abstracted later as needed.
[ bp: Massage commit message and code, squash 30-ish more fixes from
Yazen and me. ]
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Co-developed-by: <naveenkrishna.chatradhi@amd.com>
Signed-off-by: <naveenkrishna.chatradhi@amd.com>
Co-developed-by: <muralidhara.mk@amd.com>
Signed-off-by: <muralidhara.mk@amd.com>
Tested-by: <sathyapriya.k@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240214033516.1344948-3-yazen.ghannam@amd.com
Zen-based AMD systems report DRAM ECC errors through Unified Memory
Controller (UMC) MCA banks. The value provided in MCA_ADDR is
a "normalized" address which represents the UMC's view of its managed
memory. The normalized address must be translated to a system physical
address for software to take action.
MI300 systems, uniquely, do not provide a normalized address in MCA_ADDR
for DRAM ECC errors. Rather, the "DRAM" address is reported. This value
includes identifiers for the bank, row, column, pseudochannel and stack
of the memory location.
The DRAM address must be converted to a normalized address in order to
be further translated to a system physical address.
Add helper functions to do the DRAM to normalized translation for MI300
systems. The method is based on the fixed hardware layout of the on-chip
memory.
[ bp: Massage commit message, decapitalize some, rename function. ]
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>
Tested-by: Muralidhara M K <muralidhara.mk@amd.com>
Link: https://lore.kernel.org/r/20240131165732.88297-1-yazen.ghannam@amd.com
AMD MI300 systems include on-die HBM3 memory and a unique topology. And
they fall under Data Fabric version 4.5 in overall design.
Generally, topology information (IDs, etc.) is gathered from Data Fabric
registers. However, the unique topology for MI300 means that some
topology information is fixed in hardware and follows arbitrary
mappings. Furthermore, not all hardware instances are software-visible,
so register accesses must be adjusted.
Recognize and add helper functions for the new MI300 interleave modes.
Add lookup tables for fixed values where appropriate. Adjust how Die and
Node IDs are found and used.
Also, fix some register bitmasks that were mislabeled.
Signed-off-by: Muralidhara M K <muralidhara.mk@amd.com>
Co-developed-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240128155950.1434067-1-yazen.ghannam@amd.com
There are a lot of RAS topic to document, and there are a lot of details
for each topic.
Prep for this by adding an index for the RAS directory. This will
provide a top-level document and table of contents. It also provides the
option to build the RAS directory individually using "make SPHINXDIRS=".
Start a section on address translation. This will be expanded with
details for future translation methods and how they're used in the
kernel.
Move the error decoding topic to its own section. Links to other error
decoding kernel docs will be added.
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240123041401.79812-4-yazen.ghannam@amd.com
AMD Zen-based systems report memory errors through Machine Check banks
representing Unified Memory Controllers (UMCs). The address value
reported for DRAM ECC errors is a "normalized address" that is relative
to the UMC. This normalized address must be converted to a system
physical address to be usable by the OS.
Support for this address translation was introduced to the MCA subsystem
with Zen1 systems. The code was later moved to the AMD64 EDAC module,
since this was the only user of the code at the time.
However, there are uses for this translation outside of EDAC. The system
physical address can be used in MCA for preemptive page offlining as done
in some MCA notifier functions. Also, this translation is needed as the
basis of similar functionality needed for some CXL configurations on AMD
systems.
Introduce a common address translation library that can be used for
multiple subsystems including MCA, EDAC, and CXL.
Include support for UMC normalized to system physical address
translation for current CPU systems.
The Data Fabric Indirect register access offsets and one of the register
fields were changed. Default to the current offsets and register field
definition. And fallback to the older values if running on a "legacy"
system.
Provide built-in code to facilitate the loading and unloading of the
library module without affecting other modules or built-in code.
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20240123041401.79812-2-yazen.ghannam@amd.com
Pull more bcachefs updates from Kent Overstreet:
"Some fixes, Some refactoring, some minor features:
- Assorted prep work for disk space accounting rewrite
- BTREE_TRIGGER_ATOMIC: after combining our trigger callbacks, this
makes our trigger context more explicit
- A few fixes to avoid excessive transaction restarts on
multithreaded workloads: fstests (in addition to ktest tests) are
now checking slowpath counters, and that's shaking out a few bugs
- Assorted tracepoint improvements
- Starting to break up bcachefs_format.h and move on disk types so
they're with the code they belong to; this will make room to start
documenting the on disk format better.
- A few minor fixes"
* tag 'bcachefs-2024-01-21' of https://evilpiepirate.org/git/bcachefs: (46 commits)
bcachefs: Improve inode_to_text()
bcachefs: logged_ops_format.h
bcachefs: reflink_format.h
bcachefs; extents_format.h
bcachefs: ec_format.h
bcachefs: subvolume_format.h
bcachefs: snapshot_format.h
bcachefs: alloc_background_format.h
bcachefs: xattr_format.h
bcachefs: dirent_format.h
bcachefs: inode_format.h
bcachefs; quota_format.h
bcachefs: sb-counters_format.h
bcachefs: counters.c -> sb-counters.c
bcachefs: comment bch_subvolume
bcachefs: bch_snapshot::btime
bcachefs: add missing __GFP_NOWARN
bcachefs: opts->compression can now also be applied in the background
bcachefs: Prep work for variable size btree node buffers
bcachefs: grab s_umount only if snapshotting
...
Pull timer updates from Thomas Gleixner:
"Updates for time and clocksources:
- A fix for the idle and iowait time accounting vs CPU hotplug.
The time is reset on CPU hotplug which makes the accumulated
systemwide time jump backwards.
- Assorted fixes and improvements for clocksource/event drivers"
* tag 'timers-core-2024-01-21' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
tick-sched: Fix idle and iowait sleeptime accounting vs CPU hotplug
clocksource/drivers/ep93xx: Fix error handling during probe
clocksource/drivers/cadence-ttc: Fix some kernel-doc warnings
clocksource/drivers/timer-ti-dm: Fix make W=n kerneldoc warnings
clocksource/timer-riscv: Add riscv_clock_shutdown callback
dt-bindings: timer: Add StarFive JH8100 clint
dt-bindings: timer: thead,c900-aclint-mtimer: separate mtime and mtimecmp regs
Pull powerpc fixes from Aneesh Kumar:
- Increase default stack size to 32KB for Book3S
Thanks to Michael Ellerman.
* tag 'powerpc-6.8-2' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/64s: Increase default stack size to 32KB
Add a field to bch_snapshot for creation time; this will be important
when we start exposing the snapshot tree to userspace.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The "apply this compression method in the background" paths now use the
compression option if background_compression is not set; this means that
setting or changing the compression option will cause existing data to
be compressed accordingly in the background.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
bcachefs btree nodes are big - typically 256k - and btree roots are
pinned in memory. As we're now up to 18 btrees, we now have significant
memory overhead in mostly empty btree roots.
And in the future we're going to start enforcing that certain btree node
boundaries exist, to solve lock contention issues - analagous to XFS's
AGIs.
Thus, we need to start allocating smaller btree node buffers when we
can. This patch changes code that refers to the filesystem constant
c->opts.btree_node_size to refer to the btree node buffer size -
btree_buf_bytes() - where appropriate.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The variable tmp is being assigned a value but it isn't being
read afterwards. The assignment is redundant and so tmp can be
removed.
Cleans up clang scan build warning:
warning: Although the value stored to 'ret' is used in the enclosing
expression, the value is never actually read from 'ret'
[deadcode.DeadStores]
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
drop_locks_do() should not be used in a fastpath without first trying
the do in nonblocking mode - the unlock and relock will cause excessive
transaction restarts and potentially livelocking with other threads that
are contending for the same locks.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Factor out bch2_journal_bufs_to_text(), and use it in the
journal_entry_full() tracepoint; when we can't get a journal reservation
we need to know the outstanding journal entry sizes to know if the
problem is due to excessive flushing.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
When issuing discards, we may need to flush the journal if there's too
many buckets that can't be discarded until a journal flush.
But the heuristic was bad; we should be comparing the number of buckets
that need to flushes against the number of free buckets, not the number
of buckets we saw.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Also print out the data_opts, so that we can see what specifically is
being done to an extent.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This fixes a bug with rebalance IOs getting stuck with reads completed,
but writes never being issued.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
