| Age | Commit message (Collapse) | Author |
|---|
|
git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl
Pull cxl fixes from Dave Jiang:
- Fix index of Clear Event Record handles in cxl_clear_event_record()
- Fix use before init of map->reg_type in cxl_decode_regblock()
- Fix initialization of mbox_cmd.size_out in cxl_mem_get_records_log()
- Fix CXL path access_coordinate computation:
- Remove unneded check of iter in loop
- Fix of retrieving of access_coordinate in PCI topology walk
- Fix of incorrect region access_coordinate data calculation
- Consolidate of access_coordinates attached to downstream port
context
- Add check to validate access_coordinate validity to prevent
incorrect data being exposed via sysfs
* tag 'cxl-fixes-6.9-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl:
cxl: Add checks to access_coordinate calculation to fail missing data
cxl: Consolidate dport access_coordinate ->hb_coord and ->sw_coord into ->coord
cxl: Fix incorrect region perf data calculation
cxl: Fix retrieving of access_coordinates in PCIe path
cxl: Remove checking of iter in cxl_endpoint_get_perf_coordinates()
cxl/core: Fix initialization of mbox_cmd.size_out in get event
cxl/core/regs: Fix usage of map->reg_type in cxl_decode_regblock() before assigned
cxl/mem: Fix for the index of Clear Event Record Handle
|
|
Jonathan noted that when the coordinates for host bridge and switches
can be 0s if no actual data are retrieved and the calculation continues.
The resulting number would be inaccurate. Add checks to ensure that the
calculation would complete only if the numbers are valid.
While not seen in the wild, issue may show up with a BIOS that reported
CXL root ports via Generic Ports (via a PCI handle in the SRAT entry).
Fixes: 14a6960b3e92 ("cxl: Add helper function that calculate performance data for downstream ports")
Reported-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Davidlohr Bueso <dave@stgolabs.net>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/r/20240403154844.3403859-6-dave.jiang@intel.com
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
|
|
The driver stores access_coordinate for host bridge in ->hb_coord and
switch CDAT access_coordinate in ->sw_coord. Since neither of these
access_coordinate clobber each other, the variable name can be consolidated
into ->coord to simplify the code.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Davidlohr Bueso <dave@stgolabs.net>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/r/20240403154844.3403859-5-dave.jiang@intel.com
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
|
|
Current math in cxl_region_perf_data_calculate divides the latency by 1000
every time the function gets called. This causes the region latency to be
divided by 1000 per memory device and the math is incorrect. This is user
visible as the latency access_coordinate exposed via sysfs will show
incorrect latency data.
Normalize values from CDAT to nanoseconds. Adjust sub-nanoseconds latency
to at least 1. Remove adjustment of perf numbers from the generic target
since hmat handling code has already normalized those numbers. Now all
computation and stored numbers should be in nanoseconds.
cxl_hb_get_perf_coordinates() is removed and HB coords are calculated
in the port access_coordinate calculation path since it no longer need
to be treated special.
Fixes: 3d9f4a197230 ("cxl/region: Calculate performance data for a region")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/r/20240403154844.3403859-4-dave.jiang@intel.com
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
|
|
Current loop in cxl_endpoint_get_perf_coordinates() incorrectly assumes
the Root Port (RP) dport is the one with generic port access_coordinate.
However those coordinates are one level up in the Host Bridge (HB).
Current code causes the computation code to pick up 0s as the coordinates
and cause minimal bandwidth to result in 0.
Add check to skip RP when combining coordinates.
Fixes: 14a6960b3e92 ("cxl: Add helper function that calculate performance data for downstream ports")
Reported-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/r/20240403154844.3403859-3-dave.jiang@intel.com
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
|
|
The while() loop in cxl_endpoint_get_perf_coordinates() checks to see if
'iter' is valid as part of the condition breaking out of the loop.
is_cxl_root() will stop the loop before the next iteration could go NULL.
Remove the iter check.
The presence of the iter or removing the iter does not impact the behavior
of the code. This is a code clean up and not a bug fix.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Davidlohr Bueso <dave@stgolabs.net>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/r/20240403154844.3403859-2-dave.jiang@intel.com
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
|
|
Since mbox_cmd.size_out is overwritten with the actual output size in
the function below, it needs to be initialized every time.
cxl_internal_send_cmd -> __cxl_pci_mbox_send_cmd
Problem scenario:
1) The size_out variable is initially set to the size of the mailbox.
2) Read an event.
- size_out is set to 160 bytes(header 32B + one event 128B).
- Two event are created while reading.
3) Read the new *two* events.
- size_out is still set to 160 bytes.
- Although the value of out_len is 288 bytes, only 160 bytes are
copied from the mailbox register to the local variable.
- record_count is set to 2.
- Accessing records[1] will result in reading incorrect data.
Fixes: 6ebe28f9ec72 ("cxl/mem: Read, trace, and clear events on driver load")
Tested-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Kwangjin Ko <kwangjin.ko@sk.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
|
|
assigned
In the error path, map->reg_type is being used for kernel warning
before its value is setup. Found by code inspection. Exposure to
user is wrong reg_type being emitted via kernel log. Use a local
var for reg_type and retrieve value for usage.
Fixes: 6c7f4f1e51c2 ("cxl/core/regs: Make cxl_map_{component, device}_regs() device generic")
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Davidlohr Bueso <dave@stgolabs.net>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
|
|
The dev_dbg info for Clear Event Records mailbox command would report
the handle of the next record to clear not the current one.
This was because the index 'i' had incremented before printing the
current handle value.
Fixes: 6ebe28f9ec72 ("cxl/mem: Read, trace, and clear events on driver load")
Signed-off-by: Yuquan Wang <wangyuquan1236@phytium.com.cn>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
|
|
Commit 5d7107c72796 ("perf: CXL Performance Monitoring Unit driver")
added the config entries for CXL_PMU in drivers/cxl/Kconfig and
drivers/perf/Kconfig, so it can be toggled from multiple locations:
[1] Device Drivers
-> PCI support
-> CXL (Compute Expres Link) Devices
-> CXL Performance Monitoring Unit
[2] Device Drivers
-> Performance monitor support
-> CXL Performance Monitoring Unit
This complicates things, and nobody else does this.
I kept the one in drivers/perf/Kconfig because CONFIG_CXL_PMU controls
the compilation of drivers/perf/cxl_pmu.c.
Acked-by: Davidlohr Bueso <dave@stgolabs.net>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace
Pull tracing updates from Steven Rostedt:
"Main user visible change:
- User events can now have "multi formats"
The current user events have a single format. If another event is
created with a different format, it will fail to be created. That
is, once an event name is used, it cannot be used again with a
different format. This can cause issues if a library is using an
event and updates its format. An application using the older format
will prevent an application using the new library from registering
its event.
A task could also DOS another application if it knows the event
names, and it creates events with different formats.
The multi-format event is in a different name space from the single
format. Both the event name and its format are the unique
identifier. This will allow two different applications to use the
same user event name but with different payloads.
- Added support to have ftrace_dump_on_oops dump out instances and
not just the main top level tracing buffer.
Other changes:
- Add eventfs_root_inode
Only the root inode has a dentry that is static (never goes away)
and stores it upon creation. There's no reason that the thousands
of other eventfs inodes should have a pointer that never gets set
in its descriptor. Create a eventfs_root_inode desciptor that has a
eventfs_inode descriptor and a dentry pointer, and only the root
inode will use this.
- Added WARN_ON()s in eventfs
There's some conditionals remaining in eventfs that should never be
hit, but instead of removing them, add WARN_ON() around them to
make sure that they are never hit.
- Have saved_cmdlines allocation also include the map_cmdline_to_pid
array
The saved_cmdlines structure allocates a large amount of data to
hold its mappings. Within it, it has three arrays. Two are already
apart of it: map_pid_to_cmdline[] and saved_cmdlines[]. More memory
can be saved by also including the map_cmdline_to_pid[] array as
well.
- Restructure __string() and __assign_str() macros used in
TRACE_EVENT()
Dynamic strings in TRACE_EVENT() are declared with:
__string(name, source)
And assigned with:
__assign_str(name, source)
In the tracepoint callback of the event, the __string() is used to
get the size needed to allocate on the ring buffer and
__assign_str() is used to copy the string into the ring buffer.
There's a helper structure that is created in the TRACE_EVENT()
macro logic that will hold the string length and its position in
the ring buffer which is created by __string().
There are several trace events that have a function to create the
string to save. This function is executed twice. Once for
__string() and again for __assign_str(). There's no reason for
this. The helper structure could also save the string it used in
__string() and simply copy that into __assign_str() (it also
already has its length).
By using the structure to store the source string for the
assignment, it means that the second argument to __assign_str() is
no longer needed.
It will be removed in the next merge window, but for now add a
warning if the source string given to __string() is different than
the source string given to __assign_str(), as the source to
__assign_str() isn't even used and will be going away.
- Added checks to make sure that the source of __string() is also the
source of __assign_str() so that it can be safely removed in the
next merge window.
Included fixes that the above check found.
- Other minor clean ups and fixes"
* tag 'trace-v6.9-2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (34 commits)
tracing: Add __string_src() helper to help compilers not to get confused
tracing: Use strcmp() in __assign_str() WARN_ON() check
tracepoints: Use WARN() and not WARN_ON() for warnings
tracing: Use div64_u64() instead of do_div()
tracing: Support to dump instance traces by ftrace_dump_on_oops
tracing: Remove second parameter to __assign_rel_str()
tracing: Add warning if string in __assign_str() does not match __string()
tracing: Add __string_len() example
tracing: Remove __assign_str_len()
ftrace: Fix most kernel-doc warnings
tracing: Decrement the snapshot if the snapshot trigger fails to register
tracing: Fix snapshot counter going between two tracers that use it
tracing: Use EVENT_NULL_STR macro instead of open coding "(null)"
tracing: Use ? : shortcut in trace macros
tracing: Do not calculate strlen() twice for __string() fields
tracing: Rework __assign_str() and __string() to not duplicate getting the string
cxl/trace: Properly initialize cxl_poison region name
net: hns3: tracing: fix hclgevf trace event strings
drm/i915: Add missing ; to __assign_str() macros in tracepoint code
NFSD: Fix nfsd_clid_class use of __string_len() macro
...
|
|
The TP_STRUCT__entry that gets assigned the region name, or an
empty string if no region is present, is erroneously initialized
to the cxl_region pointer. It needs to be properly initialized
otherwise it's length is wrong and garbage chars can appear in
the kernel trace output: /sys/kernel/tracing/trace
The bad initialization was due in part to a naming conflict with
the parameter: struct cxl_region *region. The field 'region' is
already exposed externally as the region name, so changing that
to something logical, like 'region_name' is not an option. Instead
rename the internal only struct cxl_region to the commonly used
'cxlr'.
Impact is that tooling depending on that trace data can miss
picking up a valid event when searching by region name. The
TP_printk() output, if enabled, does emit the correct region
names in the dmesg log.
This was found during testing of the cxl-list option to report
media-errors for a region.
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Jonathan Cameron <jonathan.cameron@huawei.com>
Cc: Dave Jiang <dave.jiang@intel.com>
Cc: Vishal Verma <vishal.l.verma@intel.com>
Cc: stable@vger.kernel.org
Fixes: ddf49d57b841 ("cxl/trace: Add TRACE support for CXL media-error records")
Signed-off-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
|
|
Pick up a parsing fix for the CDAT SSLBIS structure for v6.9.
|
|
Pick up support for injecting errors via ACPI EINJ into the CXL protocol
for v6.9.
|
|
Pick up support for CXL "HMEM reporting" for v6.9, i.e. build an HMAT
from CXL CDAT and PCIe switch information.
|
|
There exist card implementations with a CDAT table using a fixed size
buffer, but with entries filled in that do not fill the whole table
length size. Then, the last entry in the CDAT table may not mark the
end of the CDAT table buffer specified by the length field in the CDAT
header. It can be shorter with trailing unused (zero'ed) data. The
actual table length is determined while reading all CDAT entries of
the table with DOE.
If the table is greater than expected (containing zero'ed trailing
data), the CDAT parser fails with:
[ 48.691717] Malformed DSMAS table length: (24:0)
[ 48.702084] [CDAT:0x00] Invalid zero length
[ 48.711460] cxl_port endpoint1: Failed to parse CDAT: -22
In addition, a check of the table buffer length is missing to prevent
an out-of-bound access then parsing the CDAT table.
Hardening code against device returning borked table. Fix that by
providing an optional buffer length argument to
acpi_parse_entries_array() that can be used by cdat_table_parse() to
propagate the buffer size down to its users to check the buffer
length. This also prevents a possible out-of-bound access mentioned.
Add a check to warn about a malformed CDAT table length.
Cc: Rafael J. Wysocki <rafael@kernel.org>
Cc: Len Brown <lenb@kernel.org>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Robert Richter <rrichter@amd.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/ZdEnopFO0Tl3t2O1@rric.localdomain
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Reading the CDAT table using DOE requires a Table Access Response
Header in addition to the CDAT entry. In current implementation this
has caused offsets with sizeof(__le32) to the actual buffers. This led
to hardly readable code and even bugs. E.g., see fix of devm_kfree()
in read_cdat_data():
commit c65efe3685f5 ("cxl/cdat: Free correct buffer on checksum error")
Rework code to avoid calculations with sizeof(__le32). Introduce
struct cdat_doe_rsp for this which contains the Table Access Response
Header and a variable payload size for various data structures
afterwards to access the CDAT table and its CDAT Data Structures
without recalculating buffer offsets.
Cc: Lukas Wunner <lukas@wunner.de>
Cc: Fan Ni <nifan.cxl@gmail.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Robert Richter <rrichter@amd.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20240216155844.406996-3-rrichter@amd.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Trivial variable rename for the DOE mailbox handle from cdat_doe to
doe_mb. The variable name cdat_doe is too ambiguous, use doe_mb that
is commonly used for the mailbox.
Signed-off-by: Robert Richter <rrichter@amd.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20240216155844.406996-2-rrichter@amd.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
The 'entry' pointer in cdat_sslbis_handler() is set to header +
sizeof(common header). However, the math missed the addition of the SSLBIS
main header. It should be header + sizeof(common header) + sizeof(*sslbis).
Use a defined struct for all the SSLBIS parts in order to avoid pointer
math errors.
The bug causes incorrect parsing of the SSLBIS table and introduces incorrect
performance values to the access_coordinates during the CXL access_coordinate
calculation path if there are CXL switches present in the topology.
The issue was found during testing of new code being added to add additional
checks for invalid CDAT values during CXL access_coordinate calculation. The
testing was done on qemu with a CXL topology including a CXL switch.
Fixes: 80aa780dda20 ("cxl: Add callback to parse the SSLBIS subtable from CDAT")
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>
Link: https://lore.kernel.org/r/20240301210948.1298075-1-dave.jiang@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Export CXL helper functions in einj-cxl.c for getting/injecting
available CXL protocol error types to sysfs under kernel/debug/cxl.
The kernel/debug/cxl/einj_types file will print the available CXL
protocol errors in the same format as the available_error_types
file provided by the einj module. The
kernel/debug/cxl/$dport_dev/einj_inject file is functionally the same
as the error_type and error_inject files provided by the EINJ module,
i.e.: writing an error type into $dport_dev/einj_inject will inject
said error type into the CXL dport represented by $dport_dev.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Ben Cheatham <Benjamin.Cheatham@amd.com>
Link: https://lore.kernel.org/r/20240311142508.31717-4-Benjamin.Cheatham@amd.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
For the numa nodes that are not created by SRAT, no memory_target is
allocated and is not managed by the HMAT_REPORTING code. Therefore
hmat_callback() memory hotplug notifier will exit early on those NUMA
nodes. The CXL memory hotplug notifier will need to call
node_set_perf_attrs() directly in order to setup the access sysfs
attributes.
In acpi_numa_init(), the last proximity domain (pxm) id created by SRAT is
stored. Add a helper function acpi_node_backed_by_real_pxm() in order to
check if a NUMA node id is defined by SRAT or created by CFMWS.
node_set_perf_attrs() symbol is exported to allow update of perf attribs
for a node. The sysfs path of
/sys/devices/system/node/nodeX/access0/initiators/* is created by
node_set_perf_attrs() for the various attributes where nodeX is matched
to the NUMA node of the CXL region.
Cc: Rafael J. Wysocki <rafael@kernel.org>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240308220055.2172956-13-dave.jiang@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
When the CXL region is formed, the driver computes the performance data
for the region. However this data is not available at the node data
collection that has been populated by the HMAT during kernel
initialization. Add a memory hotplug notifier to update the access
coordinates to the 'struct memory_target' context kept by the
HMAT_REPORTING code.
Add CXL_CALLBACK_PRI for a memory hotplug callback priority. Set the
priority number to be called before HMAT_CALLBACK_PRI. The CXL update must
happen before hmat_callback().
A new HMAT_REPORTING helper hmat_update_target_coordinates() is added in
order to allow CXL to update the memory_target access coordinates.
A new ext_updated member is added to the memory_target to indicate that
the access coordinates within the memory_target has been updated by an
external agent such as CXL. This prevents data being overwritten by the
hmat_update_target_attrs() triggered by hmat_callback().
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Rafael J. Wysocki <rafael@kernel.org>
Reviewed-by: Huang, Ying <ying.huang@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240308220055.2172956-12-dave.jiang@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Add read/write latencies and bandwidth sysfs attributes for the enabled CXL
region. The bandwidth is the aggregated bandwidth of all devices that
contribute to the CXL region. The latency is the worst latency of the
device amongst all the devices that contribute to the CXL region.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240308220055.2172956-11-dave.jiang@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Calculate and store the performance data for a CXL region. Find the worst
read and write latency for all the included ranges from each of the devices
that attributes to the region and designate that as the latency data. Sum
all the read and write bandwidth data for each of the device region and
that is the total bandwidth for the region.
The perf list is expected to be constructed before the endpoint decoders
are registered and thus there should be no early reading of the entries
from the region assemble action. The calling of the region qos calculate
function is under the protection of cxl_dpa_rwsem and will ensure that
all DPA associated work has completed.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240308220055.2172956-10-dave.jiang@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Move setting of cxlmd->endpoint to before calling add_device() on the port
device. Otherwise when referencing cxlmd->endpoint in region discovery code
that is triggered by the port driver probe function, the endpoint port
pointer is not valid.
Current code does not hit this issue yet since cxlmd->endpoint is not being
referenced during region discovery. However follow on code that does
performance calculations will.
Tested-by: Wonjae Lee <wj28.lee@samsung.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240308220055.2172956-9-dave.jiang@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Retrieve the qos_class (QTG ID) using the access coordinates from the
nearest CPU rather than the nearst initiator that may not be a CPU.
This may be the more appropriate number that applications care about.
For most cases, access0 and access1 have the same values.
Link: https://lore.kernel.org/linux-cxl/20240112113023.00006c50@Huawei.com/
Suggested-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240308220055.2172956-8-dave.jiang@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
The difference between access class 0 and access class 1 for 'struct
access_coordinate', if any, is that class 0 is for the distance from
the target to the closest initiator and that class 1 is for the distance
from the target to the closest CPU. For CXL memory, the nearest initiator
may not necessarily be a CPU node. The performance path from the CXL
endpoint to the host bridge should remain the same. However, the numbers
extracted and stored from HMAT is the difference for the two access
classes. Split out the performance numbers for the host bridge (generic
target) from the calculation of the entire path in order to allow
calculation of both access classes for a CXL region.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240308220055.2172956-7-dave.jiang@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Refactor the common code of combining coordinates in order to reduce code.
Create a new function cxl_cooordinates_combine() it combine two 'struct
access_coordinate'.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240308220055.2172956-6-dave.jiang@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
classes
Update acpi_get_genport_coordinates() to allow retrieval of both access
classes of the 'struct access_coordinate' for a generic target. The update
will allow CXL code to compute access coordinates for both access class.
Cc: Rafael J. Wysocki <rafael@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240308220055.2172956-5-dave.jiang@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
The expectation is that cxl_parse_cfwms() continues in the face the of
failure as evidenced by code like:
cxlrd = cxl_root_decoder_alloc(root_port, ways, cxl_calc_hb);
if (IS_ERR(cxlrd))
return 0;
There are other error paths in that function which mistakenly follow
idiomatic expectations and return an error when they should not. Most of
those mistakes are innocuous checks that hardly ever fail in practice.
However, a recent change succeed in making the implementation more
fragile by applying an idiomatic, but still wrong "fix" [1]. In this
failure case the kernel reports:
cxl root0: Failed to populate active decoder targets
cxl_acpi ACPI0017:00: Failed to add decode range: [mem 0x00000000-0x7fffffff flags 0x200]
...which is a real issue with that one window (to be fixed separately),
but ends up failing the entirety of cxl_acpi_probe().
Undo that recent breakage while also removing the confusion about
ignoring errors. Update all exits paths to return an error per typical
expectations and let an outer wrapper function handle dropping the
error.
Fixes: 91019b5bc7c2 ("cxl/acpi: Return 'rc' instead of '0' in cxl_parse_cfmws()") [1]
Cc: <stable@vger.kernel.org>
Cc: Breno Leitao <leitao@debian.org>
Cc: Alison Schofield <alison.schofield@intel.com>
Cc: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Pick up CXL CPER notification removal for v6.8-rc6, to return in a later
merge window.
|
|
Initial tests with the CXL CPER implementation identified that error
reports were being duplicated in the log and the trace event [1]. Then
it was discovered that the notification handler took sleeping locks
while the GHES event handling runs in spin_lock_irqsave() context [2]
While the duplicate reporting was fixed in v6.8-rc4, the fix for the
sleeping-lock-vs-atomic collision would enjoy more time to settle and
gain some test cycles. Given how late it is in the development cycle,
remove the CXL hookup for now and try again during the next merge
window.
Note that end result is that v6.8 does not emit CXL CPER payloads to the
kernel log, but this is in line with the CXL trend to move error
reporting to trace events instead of the kernel log.
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Rafael J. Wysocki <rafael@kernel.org>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Link: http://lore.kernel.org/r/20240108165855.00002f5a@Huawei.com [1]
Closes: http://lore.kernel.org/r/b963c490-2c13-4b79-bbe7-34c6568423c7@moroto.mountain [2]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
The Linux CXL subsystem is built on the assumption that HPA == SPA.
That is, the host physical address (HPA) the HDM decoder registers are
programmed with are system physical addresses (SPA).
During HDM decoder setup, the DVSEC CXL range registers (cxl-3.1,
8.1.3.8) are checked if the memory is enabled and the CXL range is in
a HPA window that is described in a CFMWS structure of the CXL host
bridge (cxl-3.1, 9.18.1.3).
Now, if the HPA is not an SPA, the CXL range does not match a CFMWS
window and the CXL memory range will be disabled then. The HDM decoder
stops working which causes system memory being disabled and further a
system hang during HDM decoder initialization, typically when a CXL
enabled kernel boots.
Prevent a system hang and do not disable the HDM decoder if the
decoder's CXL range is not found in a CFMWS window.
Note the change only fixes a hardware hang, but does not implement
HPA/SPA translation. Support for this can be added in a follow on
patch series.
Signed-off-by: Robert Richter <rrichter@amd.com>
Fixes: 34e37b4c432c ("cxl/port: Enable HDM Capability after validating DVSEC Ranges")
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20240216160113.407141-1-rrichter@amd.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Current implementation exports only to
/sys/bus/cxl/devices/.../memN/qos_class. With both ram and pmem exposed,
the second registered sysfs attribute is rejected as duplicate. It's not
possible to create qos_class under the dev_groups via the driver due to
the ram and pmem sysfs sub-directories already created by the device sysfs
groups. Move the ram and pmem qos_class to the device sysfs groups and add
a call to sysfs_update() after the perf data are validated so the
qos_class can be visible. The end results should be
/sys/bus/cxl/devices/.../memN/ram/qos_class and
/sys/bus/cxl/devices/.../memN/pmem/qos_class.
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240206190431.1810289-4-dave.jiang@intel.com
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
The passed in host bridge parameter for device_for_each_child() has
unnecessary void * type cast. Remove the type cast.
Suggested-by: Jonathan Cameron <Jonathan.Cameron@Huawei.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240206190431.1810289-3-dave.jiang@intel.com
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
cxl_dpa_perf'
In order to address the issue with being able to expose qos_class sysfs
attributes under 'ram' and 'pmem' sub-directories, the attributes must
be defined as static attributes rather than under driver->dev_groups.
To avoid implementing locking for accessing the 'struct cxl_dpa_perf`
lists, convert the list to a single 'struct cxl_dpa_perf' entry in
preparation to move the attributes to statically defined.
While theoretically a partition may have multiple qos_class via CDAT, this
has not been encountered with testing on available hardware. The code is
simplified for now to not support the complex case until a use case is
needed to support that.
Link: https://lore.kernel.org/linux-cxl/65b200ba228f_2d43c29468@dwillia2-mobl3.amr.corp.intel.com.notmuch/
Suggested-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20240206190431.1810289-2-dave.jiang@intel.com
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Autodiscovered regions can fail to assemble if they are not discovered
in HPA decode order. The user will see failure messages like:
[] cxl region0: endpoint5: HPA order violation region1
[] cxl region0: endpoint5: failed to allocate region reference
The check that is causing the failure helps the CXL driver enforce
a CXL spec mandate that decoders be committed in HPA order. The
check is needless for autodiscovered regions since their decoders
are already programmed. Trying to enforce order in the assembly of
these regions is useless because they are assembled once all their
member endpoints arrive, and there is no guarantee on the order in
which endpoints are discovered during probe.
Keep the existing check, but for autodiscovered regions, allow the
out of order assembly after a sanity check that the lesser numbered
decoder has the lesser HPA starting address.
Signed-off-by: Alison Schofield <alison.schofield@intel.com>
Tested-by: Wonjae Lee <wj28.lee@samsung.com>
Link: https://lore.kernel.org/r/3dec69ee97524ab229a20c6739272c3000b18408.1706736863.git.alison.schofield@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
In preparation for adding a new caller of cxl_region_find_decoders()
teach it to find a decoder from a cxl_endpoint_decoder structure.
Combining switch and endpoint decoder lookup in one function prevents
code duplication in call sites.
Update the existing caller.
Signed-off-by: Alison Schofield <alison.schofield@intel.com>
Tested-by: Wonjae Lee <wj28.lee@samsung.com>
Link: https://lore.kernel.org/r/79ae6d72978ef9f3ceec9722e1cb793820553c8e.1706736863.git.alison.schofield@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi
Pull EFI fixes from Ard Biesheuvel:
"The only notable change here is the patch that changes the way we deal
with spurious errors from the EFI memory attribute protocol. This will
be backported to v6.6, and is intended to ensure that we will not
paint ourselves into a corner when we tighten this further in order to
comply with MS requirements on signed EFI code.
Note that this protocol does not currently exist in x86 production
systems in the field, only in Microsoft's fork of OVMF, but it will be
mandatory for Windows logo certification for x86 PCs in the future.
- Tighten ELF relocation checks on the RISC-V EFI stub
- Give up if the new EFI memory attributes protocol fails spuriously
on x86
- Take care not to place the kernel in the lowest 16 MB of DRAM on
x86
- Omit special purpose EFI memory from memblock
- Some fixes for the CXL CPER reporting code
- Make the PE/COFF layout of mixed-mode capable images comply with a
strict interpretation of the spec"
* tag 'efi-fixes-for-v6.8-1' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi:
x86/efistub: Use 1:1 file:memory mapping for PE/COFF .compat section
cxl/trace: Remove unnecessary memcpy's
cxl/cper: Fix errant CPER prints for CXL events
efi: Don't add memblocks for soft-reserved memory
efi: runtime: Fix potential overflow of soft-reserved region size
efi/libstub: Add one kernel-doc comment
x86/efistub: Avoid placing the kernel below LOAD_PHYSICAL_ADDR
x86/efistub: Give up if memory attribute protocol returns an error
riscv/efistub: Tighten ELF relocation check
riscv/efistub: Ensure GP-relative addressing is not used
|
|
CPER events don't have UUIDs. Therefore UUIDs were removed from the
records passed to trace events and replaced with hard coded values.
As pointed out by Jonathan, the new defines for the UUIDs present a more
efficient way to assign UUID in trace records.[1]
Replace memcpy's with the use of static data.
[1] https://lore.kernel.org/all/20240108132325.00000e9c@Huawei.com/
Suggested-by: Jonathan Cameron <Jonathan.Cameron@Huawei.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
|
|
The PCI AER model is an awkward fit for CXL error handling. While the
expectation is that a PCI device can escalate to link reset to recover
from an AER event, the same reset on CXL amounts to a surprise memory
hotplug of massive amounts of memory.
At present, the CXL error handler attempts some optimistic error
handling to unbind the device from the cxl_mem driver after reaping some
RAS register values. This results in a "hopeful" attempt to unplug the
memory, but there is no guarantee that will succeed.
A subsequent AER notification after the memdev unbind event can no
longer assume the registers are mapped. Check for memdev bind before
reaping status register values to avoid crashes of the form:
BUG: unable to handle page fault for address: ffa00000195e9100
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
[...]
RIP: 0010:__cxl_handle_ras+0x30/0x110 [cxl_core]
[...]
Call Trace:
<TASK>
? __die+0x24/0x70
? page_fault_oops+0x82/0x160
? kernelmode_fixup_or_oops+0x84/0x110
? exc_page_fault+0x113/0x170
? asm_exc_page_fault+0x26/0x30
? __pfx_dpc_reset_link+0x10/0x10
? __cxl_handle_ras+0x30/0x110 [cxl_core]
? find_cxl_port+0x59/0x80 [cxl_core]
cxl_handle_rp_ras+0xbc/0xd0 [cxl_core]
cxl_error_detected+0x6c/0xf0 [cxl_core]
report_error_detected+0xc7/0x1c0
pci_walk_bus+0x73/0x90
pcie_do_recovery+0x23f/0x330
Longer term, the unbind and PCI_ERS_RESULT_DISCONNECT behavior might
need to be replaced with a new PCI_ERS_RESULT_PANIC.
Fixes: 6ac07883dbb5 ("cxl/pci: Add RCH downstream port error logging")
Cc: stable@vger.kernel.org
Suggested-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Li Ming <ming4.li@intel.com>
Link: https://lore.kernel.org/r/20240129131856.2458980-1-ming4.li@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Creating a region with 16 memory devices caused a problem. The div_u64_rem
function, used for dividing an unsigned 64-bit number by a 32-bit one,
faced an issue when SZ_256M * p->interleave_ways. The result surpassed
the maximum limit of the 32-bit divisor (4G), leading to an overflow
and a remainder of 0.
note: At this point, p->interleave_ways is 16, meaning 16 * 256M = 4G
To fix this issue, I replaced the div_u64_rem function with div64_u64_rem
and adjusted the type of the remainder.
Signed-off-by: Quanquan Cao <caoqq@fujitsu.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Fixes: 23a22cd1c98b ("cxl/region: Allocate HPA capacity to regions")
Cc: <stable@vger.kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
CXL 3.1 Section 3.1.1 states:
"A Function on a CXL device must not generate INTx messages if
that Function participates in CXL.cache protocol or CXL.mem
protocols."
The generic CXL memory driver only supports devices which use the
CXL.mem protocol. The current driver attempts to allocate MSI/MSI-X
vectors in anticipation of their need for mailbox interrupts or event
processing. However, the above requirement does not require a device to
support interrupts, only that they use MSI/MSI-X. For example, a device
may disable mailbox interrupts and either be configured for firmware
first or skip event processing and function.
Dave Larsen reported that the following Intel / Agilex card does not
support interrupts on function 0.
CXL: Intel Corporation Device 0ddb (rev 01) (prog-if 10 [CXL Memory Device (CXL 2.x)])
Rather than fail device probe if interrupts are not supported; flag that
irqs are not enabled and avoid features which require interrupts.
Emit messages appropriate for the situation to aid in debugging should
device behavior be unexpected due to a failure to allocate vectors.
Note that it is possible for a device to have host based event
processing through polling. However, the driver does not support
polling and it is not anticipated to be generally required. Leave that
functionality to a future patch if such a device comes along.
Reported-by: Dave Larsen <davelarsen58@gmail.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-and-tested-by: Davidlohr Bueso <dave@stgolabs.net>
Link: https://lore.kernel.org/r/20240117-dont-fail-irq-v2-1-f33f26b0e365@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
sprintf() is deprecated for sysfs, use preferred sysfs_emit() instead.
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>
Link: https://lore.kernel.org/r/20240112062709.2490947-1-ruansy.fnst@fujitsu.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Pick up the CPER to CXL driver integration work for v6.8. Some
additional cleanup of cper_estatus_print() messages is needed, but that
is to be handled incrementally.
|
|
If the firmware has configured CXL event support to be firmware first
the OS can process those events through CPER records. The CXL layer has
unique DPA to HPA knowledge and standard event trace parsing in place.
CPER records contain Bus, Device, Function information which can be used
to identify the PCI device which is sending the event.
Change the PCI driver registration to include registration of a CXL
CPER callback to process events through the trace subsystem.
Use new scoped based management to simplify the handling of the PCI
device object.
Tested-by: Smita-Koralahalli <Smita.KoralahalliChannabasappa@amd.com>
Reviewed-by: Smita-Koralahalli <Smita.KoralahalliChannabasappa@amd.com>
Link: https://lore.kernel.org/r/20231220-cxl-cper-v5-9-1bb8a4ca2c7a@intel.com
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
[djbw: use new pci_dev guard, flip init order]
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
The CXL CPER and event log records share everything but a UUID/GUID in
their structures.
Define a cxl_event union without the UUID/GUID to be shared between the
CPER and event log record formats. Adjust the code to use this union.
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Link: https://lore.kernel.org/r/20231220-cxl-cper-v5-6-1bb8a4ca2c7a@intel.com
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
The UEFI CXL CPER structure does not include the UUID. Now that the
UUID is passed separately to the trace event there is no need to have
the UUID in those structures.
Move UUID from the event record header to the raw structures. Adjust
cxl-test to Create dummy structures for creating test records.
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Link: https://lore.kernel.org/r/20231220-cxl-cper-v5-5-1bb8a4ca2c7a@intel.com
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
The UUID data is redundant in the known event trace types. The addition
of static defines allows the trace macros to create the UUID data inside
the trace thus removing unnecessary code.
Have well known trace events use static data to set the uuid field based
on the event type.
Suggested-by: Jonathan Cameron <Jonathan.Cameron@Huawei.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Link: https://lore.kernel.org/r/20231220-cxl-cper-v5-4-1bb8a4ca2c7a@intel.com
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
|
Dan points out in review that the cxl_test code could be made better
through the use of UUID's defines rather than being open coded.[1]
Create UUID defines and use them rather than open coding them.
Suggested-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Link: http://lore.kernel.org/r/65738d09e30e2_45e0129451@dwillia2-xfh.jf.intel.com.notmuch [1]
Link: https://lore.kernel.org/r/20231220-cxl-cper-v5-3-1bb8a4ca2c7a@intel.com
[djbw: clang-format uuid definitions]
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
|
