platform/kernel/linux-starfive.git
2 years agopowerpc/mm: Export memory_add_physaddr_to_nid() for modules
Michael Ellerman [Fri, 29 Jul 2022 06:13:55 +0000 (16:13 +1000)]
powerpc/mm: Export memory_add_physaddr_to_nid() for modules

The cxl_pmem module wants to call memory_add_physaddr_to_nid(), so
export the symbol.

Link: http://lore.kernel.org/r/87sfmkbfyg.fsf@mpe.ellerman.id.au
Fixes: 04ad63f086d1 ("cxl/region: Introduce cxl_pmem_region objects")
Reported-by: Sachin Sant <sachinp@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Tested-by: Sachin Sant <sachinp@linux.ibm.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/region: Introduce cxl_pmem_region objects
Dan Williams [Tue, 11 Jan 2022 16:06:40 +0000 (08:06 -0800)]
cxl/region: Introduce cxl_pmem_region objects

The LIBNVDIMM subsystem is a platform agnostic representation of system
NVDIMM / persistent memory resources. To date, the CXL subsystem's
interaction with LIBNVDIMM has been to register an nvdimm-bridge device
and cxl_nvdimm objects to proxy CXL capabilities into existing LIBNVDIMM
subsystem mechanics.

With regions the approach is the same. Create a new cxl_pmem_region
object to proxy CXL region details into a LIBNVDIMM definition. With
this enabling LIBNVDIMM can partition CXL persistent memory regions with
legacy namespace labels. A follow-on patch will add CXL region label and
CXL namespace label support to persist region configurations across
driver reload / system-reset events.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784340111.1758207.3036498385188290968.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/pmem: Fix offline_nvdimm_bus() to offline by bridge
Dan Williams [Sat, 11 Jun 2022 00:39:59 +0000 (17:39 -0700)]
cxl/pmem: Fix offline_nvdimm_bus() to offline by bridge

Be careful to only disable cxl_pmem objects related to a given
cxl_nvdimm_bridge. Otherwise, offline_nvdimm_bus() reaches across CXL
domains and disables more than is expected.

Fixes: 21083f51521f ("cxl/pmem: Register 'pmem' / cxl_nvdimm devices")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784339569.1758207.1557084545278004577.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/region: Add region driver boiler plate
Dan Williams [Tue, 15 Jun 2021 21:00:40 +0000 (14:00 -0700)]
cxl/region: Add region driver boiler plate

The CXL region driver is responsible for routing fully formed CXL
regions to one of libnvdimm, for persistent memory regions, device-dax
for volatile memory regions, or just act as an enumeration placeholder
if the region was setup and configuration locked by platform firmware.
In the platform-firmware-setup case the expectation is that region is
already accounted in the system memory map, i.e. already enabled as
"System RAM".

For now, just attach to CXL regions in the CXL_CONFIG_COMMIT state, and
take no further action.

Given this driver is just a small / simple router, include it in the
core rather than its own module.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-18-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/hdm: Commit decoder state to hardware
Dan Williams [Thu, 9 Jun 2022 05:56:37 +0000 (22:56 -0700)]
cxl/hdm: Commit decoder state to hardware

After all the soft validation of the region has completed, convey the
region configuration to hardware while being careful to commit decoders
in specification mandated order. In addition to programming the endpoint
decoder base-address, interleave ways and granularity, the switch
decoder target lists are also established.

While the kernel can enforce spec-mandated commit order, it can not
enforce spec-mandated reset order. For example, the kernel can't stop
someone from removing an endpoint device that is occupying decoderN in a
switch decoder where decoderN+1 is also committed. To reset decoderN,
decoderN+1 must be torn down first. That "tear down the world"
implementation is saved for a follow-on patch.

Callback operations are provided for the 'commit' and 'reset'
operations. While those callbacks may prove useful for CXL accelerators
(Type-2 devices with memory) the primary motivation is to enable a
simple way for cxl_test to intercept those operations.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784338418.1758207.14659830845389904356.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/region: Program target lists
Dan Williams [Mon, 6 Jun 2022 22:18:31 +0000 (15:18 -0700)]
cxl/region: Program target lists

Once the region's interleave geometry (ways, granularity, size) is
established and all the endpoint decoder targets are assigned, the next
phase is to program all the intermediate decoders. Specifically, each
CXL switch in the path between the endpoint and its CXL host-bridge
(including the logical switch internal to the host-bridge) needs to have
its decoders programmed and the target list order assigned.

The difficulty in this implementation lies in determining which endpoint
decoder ordering combinations are valid. Consider the cxl_test case of 2
host bridges, each of those host-bridges attached to 2 switches, and
each of those switches attached to 2 endpoints for a potential 8-way
interleave. The x2 interleave at the host-bridge level requires that all
even numbered endpoint decoder positions be located on the "left" hand
side of the topology tree, and the odd numbered positions on the other.
The endpoints that are peers on the same switch need to have a position
that can be routed with a dedicated address bit per-endpoint. See
check_last_peer() for the details.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784337827.1758207.132121746122685208.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/region: Attach endpoint decoders
Dan Williams [Tue, 7 Jun 2022 17:56:10 +0000 (10:56 -0700)]
cxl/region: Attach endpoint decoders

CXL regions (interleave sets) are made up of a set of memory devices
where each device maps a portion of the interleave with one of its
decoders (see CXL 2.0 8.2.5.12 CXL HDM Decoder Capability Structure).
As endpoint decoders are identified by a provisioning tool they can be
added to a region provided the region interleave properties are set
(way, granularity, HPA) and DPA has been assigned to the decoder.

The attach event triggers several validation checks, for example:
- is the DPA sized appropriately for the region
- is the decoder reachable via the host-bridges identified by the
  region's root decoder
- is the device already active in a different region position slot
- are there already regions with a higher HPA active on a given port
  (per CXL 2.0 8.2.5.12.20 Committing Decoder Programming)

...and the attach event affords an opportunity to collect data and
resources relevant to later programming the target lists in switch
decoders, for example:
- allocate a decoder at each cxl_port in the decode chain
- for a given switch port, how many the region's endpoints are hosted
  through the port
- how many unique targets (next hops) does a port need to map to reach
  those endpoints

The act of reconciling this information and deploying it to the decoder
configuration is saved for a follow-on patch.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784337277.1758207.4108508181328528703.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/acpi: Add a host-bridge index lookup mechanism
Dan Williams [Mon, 6 Jun 2022 20:32:01 +0000 (13:32 -0700)]
cxl/acpi: Add a host-bridge index lookup mechanism

The ACPI CXL Fixed Memory Window Structure (CFMWS) defines multiple
methods to determine which host bridge provides access to a given
endpoint relative to that device's position in the interleave. The
"Interleave Arithmetic" defines either a "standard modulo" /
round-random algorithm, or "xormap" based algorithm which can be defined
as a non-linear transform. Given that there are already more options
beyond "standard modulo" and that "xormap" may turn out to be ACPI CXL
specific, provide a callback for the region provisioning code to map
endpoint positions back to expected host bridge id (cxl_dport target).

For now just support the simple modulo math case and save the xormap for
a follow-on change.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-14-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/region: Enable the assignment of endpoint decoders to regions
Dan Williams [Sat, 4 Jun 2022 22:49:53 +0000 (15:49 -0700)]
cxl/region: Enable the assignment of endpoint decoders to regions

The region provisioning process involves allocating DPA to a set of
endpoint decoders, and HPA plus the region geometry to a region device.
Then the decoder is assigned to the region. At this point several
validation steps can be performed to validate that the decoder is
suitable to participate in the region.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reported-by: kernel test robot <lkp@intel.com>
Link: https://lore.kernel.org/r/165784336184.1758207.16403282029203949622.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/region: Allocate HPA capacity to regions
Dan Williams [Mon, 25 Apr 2022 18:43:44 +0000 (11:43 -0700)]
cxl/region: Allocate HPA capacity to regions

After a region's interleave parameters (ways and granularity) are set,
add a way for regions to allocate HPA (host physical address space) from
the free capacity in their parent root-decoder. The allocator for this
capacity reuses the 'struct resource' based allocator used for
CONFIG_DEVICE_PRIVATE.

Once the tuple of "ways, granularity, [uuid], and size" is set the
region configuration transitions to the CXL_CONFIG_INTERLEAVE_ACTIVE
state which is a precursor to allowing endpoint decoders to be added to
a region.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784335630.1758207.420216490941955417.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/region: Add interleave geometry attributes
Ben Widawsky [Mon, 25 Apr 2022 18:36:48 +0000 (11:36 -0700)]
cxl/region: Add interleave geometry attributes

Add ABI to allow the number of devices that comprise a region to be
set as well as the interleave granularity for the region.

Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
[djbw: reword changelog]
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-11-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/region: Add a 'uuid' attribute
Ben Widawsky [Thu, 27 May 2021 20:30:41 +0000 (13:30 -0700)]
cxl/region: Add a 'uuid' attribute

The process of provisioning a region involves triggering the creation of
a new region object, pouring in the configuration, and then binding that
configured object to the region driver to start its operation. For
persistent memory regions the CXL specification mandates that it
identified by a uuid. Add an ABI for userspace to specify a region's
uuid.

Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
[djbw: simplify locking]
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784334465.1758207.8224025435884752570.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/region: Add region creation support
Ben Widawsky [Tue, 8 Jun 2021 17:28:34 +0000 (10:28 -0700)]
cxl/region: Add region creation support

CXL 2.0 allows for dynamic provisioning of new memory regions (system
physical address resources like "System RAM" and "Persistent Memory").
Whereas DDR and PMEM resources are conveyed statically at boot, CXL
allows for assembling and instantiating new regions from the available
capacity of CXL memory expanders in the system.

Sysfs with an "echo $region_name > $create_region_attribute" interface
is chosen as the mechanism to initiate the provisioning process. This
was chosen over ioctl() and netlink() to keep the configuration
interface entirely in a pseudo-fs interface, and it was chosen over
configfs since, aside from this one creation event, the interface is
read-mostly. I.e. configfs supports cases where an object is designed to
be provisioned each boot, like an iSCSI storage target, and CXL region
creation is mostly for PMEM regions which are created usually once
per-lifetime of a server instance. This is an improvement over nvdimm
that pre-created "seed" devices that tended to confuse users looking to
determine which devices are active and which are idle.

Recall that the major change that CXL brings over previous persistent
memory architectures is the ability to dynamically define new regions.
Compare that to drivers like 'nfit' where the region configuration is
statically defined by platform firmware.

Regions are created as a child of a root decoder that encompasses an
address space with constraints. When created through sysfs, the root
decoder is explicit. When created from an LSA's region structure a root
decoder will possibly need to be inferred by the driver.

Upon region creation through sysfs, a vacant region is created with a
unique name. Regions have a number of attributes that must be configured
before the region can be bound to the driver where HDM decoder program
is completed.

An example of creating a new region:

- Allocate a new region name:
region=$(cat /sys/bus/cxl/devices/decoder0.0/create_pmem_region)

- Create a new region by name:
while
region=$(cat /sys/bus/cxl/devices/decoder0.0/create_pmem_region)
! echo $region > /sys/bus/cxl/devices/decoder0.0/create_pmem_region
do true; done

- Region now exists in sysfs:
stat -t /sys/bus/cxl/devices/decoder0.0/$region

- Delete the region, and name:
echo $region > /sys/bus/cxl/devices/decoder0.0/delete_region

Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784333909.1758207.794374602146306032.stgit@dwillia2-xfh.jf.intel.com
[djbw: simplify locking, reword changelog]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agoresource: Introduce alloc_free_mem_region()
Dan Williams [Fri, 20 May 2022 20:41:24 +0000 (13:41 -0700)]
resource: Introduce alloc_free_mem_region()

The core of devm_request_free_mem_region() is a helper that searches for
free space in iomem_resource and performs __request_region_locked() on
the result of that search. The policy choices of the implementation
conform to what CONFIG_DEVICE_PRIVATE users want which is memory that is
immediately marked busy, and a preference to search for the first-fit
free range in descending order from the top of the physical address
space.

CXL has a need for a similar allocator, but with the following tweaks:

1/ Search for free space in ascending order

2/ Search for free space relative to a given CXL window

3/ 'insert' rather than 'request' the new resource given downstream
   drivers from the CXL Region driver (like the pmem or dax drivers) are
   responsible for request_mem_region() when they activate the memory
   range.

Rework __request_free_mem_region() into get_free_mem_region() which
takes a set of GFR_* (Get Free Region) flags to control the allocation
policy (ascending vs descending), and "busy" policy (insert_resource()
vs request_region()).

As part of the consolidation of the legacy GFR_REQUEST_REGION case with
the new default of just inserting a new resource into the free space
some minor cleanups like not checking for NULL before calling
devres_free() (which does its own check) is included.

Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/linux-cxl/20220420143406.GY2120790@nvidia.com/
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784333333.1758207.13703329337805274043.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/mem: Enumerate port targets before adding endpoints
Dan Williams [Tue, 7 Jun 2022 17:35:39 +0000 (10:35 -0700)]
cxl/mem: Enumerate port targets before adding endpoints

The port scanning algorithm in devm_cxl_enumerate_ports() walks up the
topology and adds cxl_port objects starting from the root down to the
endpoint. When those ports are initially created they know all their
dports, but they do not know the downstream cxl_port instance that
represents the next descendant in the topology. Rework create_endpoint()
into devm_cxl_add_endpoint() that enumerates the downstream cxl_port
topology into each port's 'struct cxl_ep' record for each endpoint it
that the port is an ancestor.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-7-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/hdm: Add sysfs attributes for interleave ways + granularity
Ben Widawsky [Sun, 10 Apr 2022 22:26:13 +0000 (15:26 -0700)]
cxl/hdm: Add sysfs attributes for interleave ways + granularity

The region provisioning flow involves selecting interleave ways +
granularity settings for a region, and then programming the decoder
topology to meet those constraints, if possible. For example, root
decoders set the minimum interleave ways + granularity for any hosted
regions.

Given decoder programming is not atomic and collisions can occur between
multiple requesting regions userspace will be responsible for conflict
resolution and it needs these attributes to make those decisions.

Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784332235.1758207.7185062713652694607.stgit@dwillia2-xfh.jf.intel.com
[djbw: reword changelog, make read-only, add sysfs ABI documentaion]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/port: Move dport tracking to an xarray
Dan Williams [Sat, 28 May 2022 03:51:19 +0000 (20:51 -0700)]
cxl/port: Move dport tracking to an xarray

Reduce the complexity and the overhead of walking the topology to
determine endpoint connectivity to root decoder interleave
configurations.

Note that cxl_detach_ep(), after it determines that the last @ep has
departed and decides to delete the port, now needs to walk the dport
array with the device_lock() held to remove entries. Previously
list_splice_init() could be used atomically delete all dport entries at
once and then perform entry tear down outside the lock. There is no
list_splice_init() equivalent for the xarray.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784331647.1758207.6345820282285119339.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/port: Move 'cxl_ep' references to an xarray per port
Dan Williams [Fri, 27 May 2022 17:58:26 +0000 (10:58 -0700)]
cxl/port: Move 'cxl_ep' references to an xarray per port

In preparation for region provisioning that needs to walk the topology
by endpoints, use an xarray to record endpoint interest in a given port.
In addition to being more space and time efficient it also reduces the
complexity of the implementation by moving locking internal to the
xarray implementation. It also allows for a single cxl_ep reference to
be recorded in multiple xarrays.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-2-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/port: Record parent dport when adding ports
Dan Williams [Fri, 27 May 2022 17:57:01 +0000 (10:57 -0700)]
cxl/port: Record parent dport when adding ports

At the time that cxl_port instances are being created, cache the dport
from the parent port that points to this new child port. This will be
useful for region provisioning when walking the tree to calculate
decoder targets, and saves rewalking the dport list after the fact to
build this information.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-1-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/port: Record dport in endpoint references
Dan Williams [Fri, 27 May 2022 07:56:59 +0000 (00:56 -0700)]
cxl/port: Record dport in endpoint references

Recall that the primary role of the cxl_mem driver is to probe if the
given endpoint is connected to a CXL port topology. In that process it
walks its device ancestry to its PCI root port. If that root port is
also a CXL root port then the probe process adds cxl_port object
instances at switch in the path between to the root and the endpoint. As
those cxl_port instances are added, or if a previous enumeration
attempt already created the port, a 'struct cxl_ep' instance is
registered with that port to track the endpoints interested in that
port.

At the time the cxl_ep is registered the downstream egress path from the
port to the endpoint is known. Take the opportunity to record that
information as it will be needed for dynamic programming of decoder
targets during region provisioning.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784329944.1758207.15203961796832072116.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/hdm: Add support for allocating DPA to an endpoint decoder
Dan Williams [Tue, 24 May 2022 01:02:30 +0000 (18:02 -0700)]
cxl/hdm: Add support for allocating DPA to an endpoint decoder

The region provisioning flow will roughly follow a sequence of:

1/ Allocate DPA to a set of decoders

2/ Allocate HPA to a region

3/ Associate decoders with a region and validate that the DPA allocations
   and topologies match the parameters of the region.

For now, this change (step 1) arranges for DPA capacity to be allocated
and deleted from non-committed decoders based on the decoder's mode /
partition selection. Capacity is allocated from the lowest DPA in the
partition and any 'pmem' allocation blocks out all remaining ram
capacity in its 'skip' setting. DPA allocations are enforced in decoder
instance order. I.e. decoder N + 1 always starts at a higher DPA than
instance N, and deleting allocations must proceed from the
highest-instance allocated decoder to the lowest.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784329399.1758207.16732038126938632700.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/hdm: Track next decoder to allocate
Dan Williams [Tue, 24 May 2022 19:04:58 +0000 (12:04 -0700)]
cxl/hdm: Track next decoder to allocate

The CXL specification enforces that endpoint decoders are committed in
hw instance id order. In preparation for adding dynamic DPA allocation,
record the hw instance id in endpoint decoders, and enforce allocations
to occur in hw instance id order.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784328827.1758207.9627538529944559954.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/hdm: Add 'mode' attribute to decoder objects
Dan Williams [Mon, 23 May 2022 19:15:35 +0000 (12:15 -0700)]
cxl/hdm: Add 'mode' attribute to decoder objects

Recall that the Device Physical Address (DPA) space of a CXL Memory
Expander is potentially partitioned into a volatile and persistent
portion. A decoder maps a Host Physical Address (HPA) range to a DPA
range and that translation depends on the value of all previous (lower
instance number) decoders before the current one.

In preparation for allowing dynamic provisioning of regions, decoders
need an ABI to indicate which DPA partition a decoder targets. This ABI
needs to be prepared for the possibility that some other agent committed
and locked a decoder that spans the partition boundary.

Add 'decoderX.Y/mode' to endpoint decoders that indicates which
partition 'ram' / 'pmem' the decoder targets, or 'mixed' if the decoder
currently spans the partition boundary.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603881967.551046.6007594190951596439.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/hdm: Enumerate allocated DPA
Dan Williams [Fri, 22 Jul 2022 00:19:12 +0000 (17:19 -0700)]
cxl/hdm: Enumerate allocated DPA

In preparation for provisioning CXL regions, add accounting for the DPA
space consumed by existing regions / decoders. Recall, a CXL region is a
memory range comprised from one or more endpoint devices contributing a
mapping of their DPA into HPA space through a decoder.

Record the DPA ranges covered by committed decoders at initial probe of
endpoint ports relative to a per-device resource tree of the DPA type
(pmem or volatile-ram).

The cxl_dpa_rwsem semaphore is introduced to globally synchronize DPA
state across all endpoints and their decoders at once. The vast majority
of DPA operations are reads as region creation is expected to be as rare
as disk partitioning and volume creation. The device_lock() for this
synchronization is specifically avoided for concern of entangling with
sysfs attribute removal.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784327682.1758207.7914919426043855876.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/core: Define a 'struct cxl_endpoint_decoder'
Dan Williams [Sat, 21 May 2022 23:24:14 +0000 (16:24 -0700)]
cxl/core: Define a 'struct cxl_endpoint_decoder'

Previously the target routing specifics of switch decoders and platform
CXL window resource tracking of root decoders were factored out of
'struct cxl_decoder'. While switch decoders translate from SPA to
downstream ports, endpoint decoders translate from SPA to DPA.

This patch, 3 of 3, adds a 'struct cxl_endpoint_decoder' that tracks an
endpoint-specific Device Physical Address (DPA) resource. For now this
just defines ->dpa_res, a follow-on patch will handle requesting DPA
resource ranges from a device-DPA resource tree.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784327088.1758207.15502834501671201192.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/core: Define a 'struct cxl_root_decoder'
Dan Williams [Wed, 13 Jul 2022 01:38:26 +0000 (18:38 -0700)]
cxl/core: Define a 'struct cxl_root_decoder'

Previously the target routing specifics of switch decoders were factored
out of 'struct cxl_decoder' into 'struct cxl_switch_decoder'.

This patch, 2 of 3, adds a 'struct cxl_root_decoder' as a superset of a
switch decoder that also track the associated CXL window platform
resource.

Note that the reason the resource for a given root decoder needs to be
looked up after the fact (i.e. after cxl_parse_cfmws() and
add_cxl_resource()) is because add_cxl_resource() may have merged CXL
windows in order to keep them at the top of the resource tree / decode
hierarchy.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784326541.1758207.9915663937394448341.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/acpi: Track CXL resources in iomem_resource
Dan Williams [Wed, 13 Jul 2022 01:37:54 +0000 (18:37 -0700)]
cxl/acpi: Track CXL resources in iomem_resource

Recall that CXL capable address ranges, on ACPI platforms, are published
in the CEDT.CFMWS (CXL Early Discovery Table: CXL Fixed Memory Window
Structures). These windows represent both the actively mapped capacity
and the potential address space that can be dynamically assigned to a
new CXL decode configuration (region / interleave-set).

CXL endpoints like DDR DIMMs can be mapped at any physical address
including 0 and legacy ranges.

There is an expectation and requirement that the /proc/iomem interface
and the iomem_resource tree in the kernel reflect the full set of
platform address ranges. I.e. that every address range that platform
firmware and bus drivers enumerate be reflected as an iomem_resource
entry. The hard requirement to do this for CXL arises from the fact that
facilities like CONFIG_DEVICE_PRIVATE expect to be able to treat empty
iomem_resource ranges as free for software to use as proxy address
space. Without CXL publishing its potential address ranges in
iomem_resource, the CONFIG_DEVICE_PRIVATE mechanism may inadvertently
steal capacity reserved for runtime provisioning of new CXL regions.

So, iomem_resource needs to know about both active and potential CXL
resource ranges. The active CXL resources might already be reflected in
iomem_resource as "System RAM". insert_resource_expand_to_fit() handles
re-parenting "System RAM" underneath a CXL window.

The "_expand_to_fit()" behavior handles cases where a CXL window is not
a strict superset of an existing entry in the iomem_resource tree. The
"_expand_to_fit()" behavior is acceptable from the perspective of
resource allocation. The expansion happens because a conflicting
resource range is already populated, which means the resource boundary
expansion does not result in any additional free CXL address space being
made available. CXL address space allocation is always bounded by the
orginal unexpanded address range.

However, the potential for expansion does mean that something like
walk_iomem_res_desc(IORES_DESC_CXL...) can only return fuzzy answers on
corner case platforms that cause the resource tree to expand a CXL
window resource over a range that is not decoded by CXL. This would be
an odd platform configuration, but if it becomes a problem in practice
the CXL subsytem could just publish an API that returns definitive
answers.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Link: https://lore.kernel.org/r/165784325943.1758207.5310344844375305118.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/core: Define a 'struct cxl_switch_decoder'
Dan Williams [Thu, 19 May 2022 00:52:23 +0000 (17:52 -0700)]
cxl/core: Define a 'struct cxl_switch_decoder'

Currently 'struct cxl_decoder' contains the superset of attributes
needed for all decoder types. Before more type-specific attributes are
added to the common definition, reorganize 'struct cxl_decoder' into type
specific objects.

This patch, the first of three, factors out a cxl_switch_decoder type.
See the new kdoc for what a 'struct cxl_switch_decoder' represents in a
CXL topology.

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reported-by: kernel test robot <lkp@intel.com>
Link: https://lore.kernel.org/r/165784325340.1758207.5064717153608954960.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agoDocumentation/cxl: Use a double line break between entries
Dan Williams [Sun, 10 Jul 2022 16:56:05 +0000 (09:56 -0700)]
Documentation/cxl: Use a double line break between entries

Make it easier to read delineations between the "Description" line
break, new paragraph line breaks, and new entries.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165784324750.1758207.10379257962719807754.stgit@dwillia2-xfh.jf.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/port: Read CDAT table
Ira Weiny [Tue, 19 Jul 2022 20:52:49 +0000 (13:52 -0700)]
cxl/port: Read CDAT table

The per-device CDAT data provides performance data that is relevant for
mapping which CXL devices can participate in which CXL ranges by QTG
(QoS Throttling Group) (per ECN: CXL 2.0 CEDT CFMWS & QTG_DSM) [1]. The
QTG association specified in the ECN is advisory. Until the
cxl_acpi driver grows support for invoking the QTG _DSM method the CDAT
data is only of interest to userspace that may need it for debug
purposes.

Search the DOE mailboxes available, query CDAT data, cache the data and
make it available via a sysfs binary attribute per endpoint at:

/sys/bus/cxl/devices/endpointX/CDAT

...similar to other ACPI-structured table data in
/sys/firmware/ACPI/tables. The CDAT is relative to 'struct cxl_port'
objects since switches in addition to endpoints can host a CDAT
instance. Switch CDAT support is not implemented.

This does not support table updates at runtime. It will always provide
whatever was there when first cached. It is also the case that table
updates are not expected outside of explicit DPA address map affecting
commands like Set Partition with the immediate flag set. Given that the
driver does not support Set Partition with the immediate flag set there
is no current need for update support.

Link: https://www.computeexpresslink.org/spec-landing
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Co-developed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
[djbw: drop in-kernel parsing infra for now, and other minor fixups]
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220719205249.566684-7-ira.weiny@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agodriver-core: Introduce BIN_ATTR_ADMIN_{RO,RW}
Ira Weiny [Tue, 19 Jul 2022 20:52:48 +0000 (13:52 -0700)]
driver-core: Introduce BIN_ATTR_ADMIN_{RO,RW}

Many binary attributes need to limit access to CAP_SYS_ADMIN only; ie
many binary attributes specify is_visible with 0400 or 0600.

Make setting the permissions of such attributes more explicit by
defining BIN_ATTR_ADMIN_{RO,RW}.

Cc: Bjorn Helgaas <bhelgaas@google.com>
Suggested-by: Dan Williams <dan.j.williams@intel.com>
Suggested-by: Krzysztof Wilczyński <kw@linux.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Link: https://lore.kernel.org/r/20220719205249.566684-6-ira.weiny@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/pci: Create PCI DOE mailbox's for memory devices
Ira Weiny [Tue, 19 Jul 2022 20:52:47 +0000 (13:52 -0700)]
cxl/pci: Create PCI DOE mailbox's for memory devices

DOE mailbox objects will be needed for various mailbox communications
with each memory device.

Iterate each DOE mailbox capability and create PCI DOE mailbox objects
as found.

It is not anticipated that this is the final resting place for the
iteration of the DOE devices.  The support of switch ports will drive
this code into the PCIe side.  In this imagined architecture the CXL
port driver would then query into the PCI device for the DOE mailbox
array.

For now creating the mailboxes in the CXL port is good enough for the
endpoints.  Later PCIe ports will need to support this to support switch
ports more generically.

Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Lukas Wunner <lukas@wunner.de>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Link: https://lore.kernel.org/r/20220719205249.566684-5-ira.weiny@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agoPCI/DOE: Add DOE mailbox support functions
Jonathan Cameron [Tue, 19 Jul 2022 20:52:46 +0000 (13:52 -0700)]
PCI/DOE: Add DOE mailbox support functions

Introduced in a PCIe r6.0, sec 6.30, DOE provides a config space based
mailbox with standard protocol discovery.  Each mailbox is accessed
through a DOE Extended Capability.

Each DOE mailbox must support the DOE discovery protocol in addition to
any number of additional protocols.

Define core PCIe functionality to manage a single PCIe DOE mailbox at a
defined config space offset.  Functionality includes iterating,
creating, query of supported protocol, and task submission.  Destruction
of the mailboxes is device managed.

Cc: "Li, Ming" <ming4.li@intel.com>
Cc: Bjorn Helgaas <helgaas@kernel.org>
Cc: Matthew Wilcox <willy@infradead.org>
Acked-by: Bjorn Helgaas <helgaas@kernel.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Co-developed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Link: https://lore.kernel.org/r/20220719205249.566684-4-ira.weiny@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agoPCI: Replace magic constant for PCI Sig Vendor ID
Ira Weiny [Tue, 19 Jul 2022 20:52:45 +0000 (13:52 -0700)]
PCI: Replace magic constant for PCI Sig Vendor ID

Replace the magic value in pci_bus_crs_vendor_id() with
PCI_VENDOR_ID_PCI_SIG.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Davidlohr Bueso <dave@stgolabs.net>
Suggested-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Ira Weiny <ira.weiny@intel.com>
Link: https://lore.kernel.org/r/20220719205249.566684-3-ira.weiny@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agoPCI: Add vendor ID for the PCI SIG
Jonathan Cameron [Tue, 19 Jul 2022 20:52:44 +0000 (13:52 -0700)]
PCI: Add vendor ID for the PCI SIG

This ID is used in DOE headers to identify protocols that are defined
within the PCI Express Base Specification, PCIe r6.0, sec 6.30.1.1 table
6-32.

Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Davidlohr Bueso <dave@stgolabs.net>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220719205249.566684-2-ira.weiny@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/pmem: Delete unused nvdimm attribute
Dan Williams [Thu, 23 Jun 2022 21:58:45 +0000 (14:58 -0700)]
cxl/pmem: Delete unused nvdimm attribute

While there is a need to go from a LIBNVDIMM 'struct nvdimm' to a CXL
'struct cxl_nvdimm', there is no use case to go the other direction.
Likely this is a leftover from an early version of the referenced commit
before it implemented devm for releasing the created nvdimm.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-19-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/hdm: Initialize decoder type for memory expander devices
Dan Williams [Tue, 7 Jun 2022 18:21:40 +0000 (11:21 -0700)]
cxl/hdm: Initialize decoder type for memory expander devices

Unless and until accelerator (type-2) drivers start registering for
CXL.mem mapping services from the CXL subsystem core, initialize idle
HDM decoders to the "expander" type. I.e. the only CXL devices using the
CXL core presently are those implementing the CXL 2.0 Type-3 memory
expander device class code that the cxl_pci driver claims.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-6-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/port: Cache CXL host bridge data
Dan Williams [Wed, 1 Jun 2022 19:49:32 +0000 (12:49 -0700)]
cxl/port: Cache CXL host bridge data

Region creation has need for checking host-bridge connectivity when
adding endpoints to regions. Record, at port creation time, the
host-bridge to provide a useful shortcut from any location in the
topology to the most-significant ancestor.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/20220624041950.559155-4-dan.j.williams@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agotools/testing/cxl: Fix decoder default state
Dan Williams [Tue, 24 May 2022 17:48:59 +0000 (10:48 -0700)]
tools/testing/cxl: Fix decoder default state

The 'enabled' state is reserved for committed decoders. By default,
cxl_test decoders are uncommitted at init time.

Fixes: 7c7d68db0254 ("tools/testing/cxl: Enumerate mock decoders")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603888091.551046.6312322707378021172.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agotools/testing/cxl: Add partition support
Dan Williams [Tue, 24 May 2022 06:26:11 +0000 (23:26 -0700)]
tools/testing/cxl: Add partition support

In support of testing DPA allocation mechanisms in the CXL core, the
cxl_test environment needs to support establishing and retrieving the
'pmem partition boundary.

Replace the platform_device_add_resources() method for delineating DPA
within an endpoint with an emulated DEV_SIZE amount of partitionable
capacity. Set DEV_SIZE such that an endpoint has enough capacity to
simultaneously participate in 8 distinct regions.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603887411.551046.13234212587991192347.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agotools/testing/cxl: Expand CFMWS windows
Dan Williams [Tue, 24 May 2022 15:56:58 +0000 (08:56 -0700)]
tools/testing/cxl: Expand CFMWS windows

For the x2 host-bridge interleave windows, allow for a
x8-endpoint-interleave configuration per memory-type with each device
contributing the minimum 256MB extent. Similarly, for the x1 host-bridge
interleave windows, allow for a x4-endpoint-interleave configuration per
memory-type.

Bump up the number of decoders per-port to support hosting 8 regions.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603886721.551046.8682583835505795210.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agotools/testing/cxl: Move cxl_test resources to the top of memory
Dan Williams [Thu, 23 Jun 2022 01:02:37 +0000 (18:02 -0700)]
tools/testing/cxl: Move cxl_test resources to the top of memory

A recent QEMU upgrade resulted in collisions between QEMU's chosen
location for PCI MMIO and cxl_test's fake address location for emulated
CXL purposes. This was great for testing resource collisions, but not so
great for continuing to test the nominal cases. Move cxl_test to the
top-of-memory where it is less likely to collide with other resources.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603886021.551046.12395967874222763381.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/mem: Add a debugfs version of 'iomem' for DPA, 'dpamem'
Dan Williams [Thu, 26 May 2022 19:15:25 +0000 (12:15 -0700)]
cxl/mem: Add a debugfs version of 'iomem' for DPA, 'dpamem'

Dump the device-physical-address map for a CXL expander in /proc/iomem
style format. E.g.:

  cat /sys/kernel/debug/cxl/mem1/dpamem
  00000000-0fffffff : ram
  10000000-1fffffff : pmem

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603885318.551046.8308248564880066726.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/debug: Move debugfs init to cxl_core_init()
Dan Williams [Sun, 10 Jul 2022 16:57:28 +0000 (09:57 -0700)]
cxl/debug: Move debugfs init to cxl_core_init()

In preparation for a new cxl debugfs file, move 'cxl' directory
establishment and teardown to the core and let subsequent init routines
reference that setup.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603884654.551046.4962104601691723080.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/Documentation: List attribute permissions
Dan Williams [Wed, 22 Jun 2022 00:23:16 +0000 (17:23 -0700)]
cxl/Documentation: List attribute permissions

Clarify the access permission of CXL sysfs attributes in the
documentation to help development of userspace tooling.

Reported-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603881198.551046.12893348287451903699.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/hdm: Require all decoders to be enumerated
Ben Widawsky [Wed, 13 Apr 2022 05:18:09 +0000 (22:18 -0700)]
cxl/hdm: Require all decoders to be enumerated

In preparation for region provisioning all device decoders need to be
enumerated since DPA allocations are calculated by summing the
capacities of all decoders in a set. I.e. the programming for decoder[N]
depends on the state of decoder[N-1], so skipping over decoders that
fail to initialize prevents accurate DPA accounting.

Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
[djbw: reword changelog]
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603879664.551046.6863805202478861026.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/mem: Convert partition-info to resources
Dan Williams [Sat, 21 May 2022 22:35:29 +0000 (15:35 -0700)]
cxl/mem: Convert partition-info to resources

To date the per-device-partition DPA range information has only been
used for enumeration purposes. In preparation for allocating regions
from available DPA capacity, convert those ranges into DPA-type resource
trees.

With resources and the new add_dpa_res() helper some open coded end
address calculations and debug prints can be cleaned.

The 'cxlds->pmem_res' and 'cxlds->ram_res' resources are child resources
of the total-device DPA space and they in turn will host DPA allocations
from cxl_endpoint_decoder instances (tracked by cxled->dpa_res).

Cc: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603878921.551046.8127845916514734142.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl: Introduce cxl_to_{ways,granularity}
Dan Williams [Mon, 23 May 2022 00:04:27 +0000 (17:04 -0700)]
cxl: Introduce cxl_to_{ways,granularity}

Interleave granularity and ways have CXL specification defined encodings.
Promote the conversion helpers to a common header, and use them to
replace other open-coded instances.

Force caller to consider the error case of the conversion similarly to
other conversion helpers like kstrto*().

Co-developed-by: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Link: https://lore.kernel.org/r/165603875016.551046.17236943065932132355.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/core: Drop is_cxl_decoder()
Dan Williams [Thu, 19 May 2022 04:38:29 +0000 (21:38 -0700)]
cxl/core: Drop is_cxl_decoder()

This helper was only used to identify the object type for lockdep
purposes. Now that lockdep support is done with explicit lock classes,
this helper can be dropped.

Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Adam Manzanares <a.manzanares@samsung.com>
Link: https://lore.kernel.org/r/165603874340.551046.15491766127759244728.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/core: Drop ->platform_res attribute for root decoders
Dan Williams [Thu, 19 May 2022 01:02:39 +0000 (18:02 -0700)]
cxl/core: Drop ->platform_res attribute for root decoders

Root decoders are responsible for hosting the available host address
space for endpoints and regions to claim. The tracking of that available
capacity can be done in iomem_resource directly. As a result, root
decoders no longer need to host their own resource tree. The
current ->platform_res attribute was added prematurely.

Otherwise, ->hpa_range fills the role of conveying the current decode
range of the decoder.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Adam Manzanares <a.manzanares@samsung.com>
Link: https://lore.kernel.org/r/165603873619.551046.791596854070136223.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/core: Rename ->decoder_range ->hpa_range
Dan Williams [Thu, 19 May 2022 00:55:13 +0000 (17:55 -0700)]
cxl/core: Rename ->decoder_range ->hpa_range

In preparation for growing a ->dpa_range attribute for endpoint
decoders, rename the current ->decoder_range to the more descriptive
->hpa_range.

Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Adam Manzanares <a.manzanares@samsung.com>
Link: https://lore.kernel.org/r/165603872867.551046.2170426227407458814.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/hdm: Use local hdm variable
Ben Widawsky [Thu, 28 Apr 2022 18:15:40 +0000 (11:15 -0700)]
cxl/hdm: Use local hdm variable

Save a few characters and use the already initialized local variable.

Signed-off-by: Ben Widawsky <bwidawsk@kernel.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Adam Manzanares <a.manzanares@samsung.com>
Link: https://lore.kernel.org/r/165603872171.551046.913207574344536475.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/port: Keep port->uport valid for the entire life of a port
Dan Williams [Fri, 3 Jun 2022 23:43:48 +0000 (16:43 -0700)]
cxl/port: Keep port->uport valid for the entire life of a port

The upcoming region provisioning implementation has a need to
dereference port->uport during the port unregister flow. Specifically,
endpoint decoders need to be able to lookup their corresponding memdev
via port->uport.

The existing ->dead flag was added for cases where the core was
committed to tearing down the port, but needed to drop locks before
calling device_unregister(). Reuse that flag to indicate to
delete_endpoint() that it has no "release action" work to do as
unregister_port() will handle it.

Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Reviewed-by: Adam Manzanares <a.manzanares@samsung.com>
Link: https://lore.kernel.org/r/165603871491.551046.6682199179541194356.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agotools/testing/cxl: Fix cxl_hdm_decode_init() calling convention
Dan Williams [Wed, 22 Jun 2022 08:00:45 +0000 (01:00 -0700)]
tools/testing/cxl: Fix cxl_hdm_decode_init() calling convention

This failing signature:

[    8.392669] cxl_bus_probe: cxl_port endpoint2: probe: 970997760
[    8.392670] cxl_port: probe of endpoint2 failed with error 970997760
[    8.392719] create_endpoint: cxl_mem mem0: add: endpoint2
[    8.392721] cxl_mem mem0: endpoint2 failed probe
[    8.392725] cxl_bus_probe: cxl_mem mem0: probe: -6

...shows cxl_hdm_decode_init() resulting in a return code ("970997760")
that looks like stack corruption. The problem goes away if
cxl_hdm_decode_init() is not mocked via __wrap_cxl_hdm_decode_init().

The corruption results from the mismatch that the calling convention for
cxl_hdm_decode_init() is:

int cxl_hdm_decode_init(struct cxl_dev_state *cxlds, struct cxl_hdm *cxlhdm)

...and __wrap_cxl_hdm_decode_init() is:

bool __wrap_cxl_hdm_decode_init(struct cxl_dev_state *cxlds, struct cxl_hdm *cxlhdm)

...i.e. an int is expected but __wrap_hdm_decode_init() returns bool.

Fix the convention and cleanup the organization to match
__wrap_cxl_await_media_ready() as the difference was a red herring that
distracted from finding the bug.

Fixes: 92804edb11f0 ("cxl/pci: Drop @info argument to cxl_hdm_decode_init()")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Adam Manzanares <a.manzanares@samsung.com>
Link: https://lore.kernel.org/r/165603870776.551046.8709990108936497723.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/mbox: Fix missing variable payload checks in cmd size validation
Vishal Verma [Tue, 28 Jun 2022 22:01:09 +0000 (16:01 -0600)]
cxl/mbox: Fix missing variable payload checks in cmd size validation

The conversion of command sizes to unsigned missed a couple of checks
against variable size payloads during command validation, which made all
variable payload commands unconditionally fail. Add the checks back using
the new CXL_VARIABLE_PAYLOAD scheme.

Fixes: 26f89535a5bb ("cxl/mbox: Use type __u32 for mailbox payload sizes")
Cc: <stable@vger.kernel.org>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Alison Schofield <alison.schofield@intel.com>
Reported-by: Abhi Cs <abhi.cs@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
Link: https://lore.kernel.org/r/20220628220109.633564-1-vishal.l.verma@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agomemregion: Fix memregion_free() fallback definition
Dan Williams [Thu, 23 Jun 2022 20:02:31 +0000 (13:02 -0700)]
memregion: Fix memregion_free() fallback definition

In the CONFIG_MEMREGION=n case, memregion_free() is meant to be a static
inline. 0day reports:

    In file included from drivers/cxl/core/port.c:4:
    include/linux/memregion.h:19:6: warning: no previous prototype for
    function 'memregion_free' [-Wmissing-prototypes]

Mark memregion_free() static.

Fixes: 33dd70752cd7 ("lib: Uplevel the pmem "region" ida to a global allocator")
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Link: https://lore.kernel.org/r/165601455171.4042645.3350844271068713515.stgit@dwillia2-xfh
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/mbox: Use __le32 in get,set_lsa mailbox structures
Alison Schofield [Fri, 25 Feb 2022 22:14:56 +0000 (14:14 -0800)]
cxl/mbox: Use __le32 in get,set_lsa mailbox structures

CXL specification defines these as little endian.

Fixes: 60b8f17215de ("cxl/pmem: Translate NVDIMM label commands to CXL label commands")
Reported-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Alison Schofield <alison.schofield@intel.com>
Link: https://lore.kernel.org/r/20220225221456.1025635-1-alison.schofield@intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl/core: Use is_endpoint_decoder
Ben Widawsky [Fri, 4 Mar 2022 21:36:45 +0000 (13:36 -0800)]
cxl/core: Use is_endpoint_decoder

Save some characters and directly check decoder type rather than port
type. There's no need to check if the port is an endpoint port since, by
this point, cxl_endpoint_decoder_alloc() has a specified type.

Reviewed by: Adam Manzanares <a.manzanares@samsung.com>
Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agocxl: Fix cleanup of port devices on failure to probe driver.
Jonathan Cameron [Thu, 9 Jun 2022 13:45:19 +0000 (14:45 +0100)]
cxl: Fix cleanup of port devices on failure to probe driver.

The device is created, and then there is a check if a driver succesfully
bound to it. In event of failing the bind (e.g. failure in cxl_port_probe())
the device is left registered. When a bus rescan later occurs, fresh
devices are created leading to a multiple device representing the same
underlying hardware. Bad things may follow and at very least we have far too many
devices.

Fix by ensuring autoremove is registered if the device create succeeds,
but doesn't depend on sucessful binding to a driver.

Bug was observed as side effect of incorrect ownership in
[PATCH v9 6/9] cxl/port: Read CDAT table
but will result from any failure to in cxl_port_probe().

Fixes: 8dd2bc0f8e02 ("cxl/mem: Add the cxl_mem driver")
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Link: https://lore.kernel.org/r/20220609134519.11668-1-Jonathan.Cameron@huawei.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agoMAINTAINERS: Update Ben's email address
Ben Widawsky [Fri, 20 May 2022 18:00:59 +0000 (11:00 -0700)]
MAINTAINERS: Update Ben's email address

I'm leaving Intel. Update email address to korg and add .mailmap
entries. For now, I will be taking a reduced role in CXL development,
but I still plan to spend time working on it, and I can still serve as a
good substitute if needed for maintainer responsibilities (that may
change in the future).

Signed-off-by: Ben Widawsky <ben.widawsky@intel.com>
Reviewed-by: Alison Schofield <alison.schofield@intel.com>
Link: https://lore.kernel.org/r/20220520180059.632054-1-bwidawsk@kernel.org
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
2 years agoLinux 5.19-rc3
Linus Torvalds [Sun, 19 Jun 2022 20:06:47 +0000 (15:06 -0500)]
Linux 5.19-rc3

2 years agoMerge tag 'x86-urgent-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 19 Jun 2022 14:58:28 +0000 (09:58 -0500)]
Merge tag 'x86-urgent-2022-06-19' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Thomas Gleixner:

 - Make RESERVE_BRK() work again with older binutils. The recent
   'simplification' broke that.

 - Make early #VE handling increment RIP when successful.

 - Make the #VE code consistent vs. the RIP adjustments and add
   comments.

 - Handle load_unaligned_zeropad() across page boundaries correctly in
   #VE when the second page is shared.

* tag 'x86-urgent-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/tdx: Handle load_unaligned_zeropad() page-cross to a shared page
  x86/tdx: Clarify RIP adjustments in #VE handler
  x86/tdx: Fix early #VE handling
  x86/mm: Fix RESERVE_BRK() for older binutils

2 years agoMerge tag 'objtool-urgent-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 19 Jun 2022 14:54:16 +0000 (09:54 -0500)]
Merge tag 'objtool-urgent-2022-06-19' of git://git./linux/kernel/git/tip/tip

Pull build tooling updates from Thomas Gleixner:

 - Remove obsolete CONFIG_X86_SMAP reference from objtool

 - Fix overlapping text section failures in faddr2line for real

 - Remove OBJECT_FILES_NON_STANDARD usage from x86 ftrace and replace it
   with finegrained annotations so objtool can validate that code
   correctly.

* tag 'objtool-urgent-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/ftrace: Remove OBJECT_FILES_NON_STANDARD usage
  faddr2line: Fix overlapping text section failures, the sequel
  objtool: Fix obsolete reference to CONFIG_X86_SMAP

2 years agoMerge tag 'sched-urgent-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 19 Jun 2022 14:51:00 +0000 (09:51 -0500)]
Merge tag 'sched-urgent-2022-06-19' of git://git./linux/kernel/git/tip/tip

Pull scheduler fix from Thomas Gleixner:
 "A single scheduler fix plugging a race between sched_setscheduler()
  and balance_push().

  sched_setscheduler() spliced the balance callbacks accross a lock
  break which makes it possible for an interleaving schedule() to
  observe an empty list"

* tag 'sched-urgent-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched: Fix balance_push() vs __sched_setscheduler()

2 years agoMerge tag 'locking-urgent-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 19 Jun 2022 14:47:41 +0000 (09:47 -0500)]
Merge tag 'locking-urgent-2022-06-19' of git://git./linux/kernel/git/tip/tip

Pull lockdep fix from Thomas Gleixner:
 "A RT fix for lockdep.

  lockdep invokes prandom_u32() to create cookies. This worked until
  prandom_u32() was switched to the real random generator, which takes a
  spinlock for extraction, which does not work on RT when invoked from
  atomic contexts.

  lockdep has no requirement for real random numbers and it turns out
  sched_clock() is good enough to create the cookie. That works
  everywhere and is faster"

* tag 'locking-urgent-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  locking/lockdep: Use sched_clock() for random numbers

2 years agoMerge tag 'irq-urgent-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 19 Jun 2022 14:45:16 +0000 (09:45 -0500)]
Merge tag 'irq-urgent-2022-06-19' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Thomas Gleixner:
 "A set of interrupt subsystem updates:

  Core:

   - Ensure runtime power management for chained interrupts

  Drivers:

   - A collection of OF node refcount fixes

   - Unbreak MIPS uniprocessor builds

   - Fix xilinx interrupt controller Kconfig dependencies

   - Add a missing compatible string to the Uniphier driver"

* tag 'irq-urgent-2022-06-19' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/loongson-liointc: Use architecture register to get coreid
  irqchip/uniphier-aidet: Add compatible string for NX1 SoC
  dt-bindings: interrupt-controller/uniphier-aidet: Add bindings for NX1 SoC
  irqchip/realtek-rtl: Fix refcount leak in map_interrupts
  irqchip/gic-v3: Fix refcount leak in gic_populate_ppi_partitions
  irqchip/gic-v3: Fix error handling in gic_populate_ppi_partitions
  irqchip/apple-aic: Fix refcount leak in aic_of_ic_init
  irqchip/apple-aic: Fix refcount leak in build_fiq_affinity
  irqchip/gic/realview: Fix refcount leak in realview_gic_of_init
  irqchip/xilinx: Remove microblaze+zynq dependency
  genirq: PM: Use runtime PM for chained interrupts

2 years agoMerge tag 'char-misc-5.19-rc3-take2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 19 Jun 2022 14:37:29 +0000 (09:37 -0500)]
Merge tag 'char-misc-5.19-rc3-take2' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes for real from Greg KH:
 "Let's tag the proper branch this time...

  Here are some small char/misc driver fixes for 5.19-rc3 that resolve
  some reported issues.

  They include:

   - mei driver fixes

   - comedi driver fix

   - rtsx build warning fix

   - fsl-mc-bus driver fix

  All of these have been in linux-next for a while with no reported
  issues"

This is what the merge in commit f0ec9c65a8d6 _should_ have merged, but
Greg fat-fingered the pull request and I got some small changes from
linux-next instead there. Credit to Nathan Chancellor for eagle-eyes.

Link: https://lore.kernel.org/all/Yqywy+Md2AfGDu8v@dev-arch.thelio-3990X/
* tag 'char-misc-5.19-rc3-take2' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  bus: fsl-mc-bus: fix KASAN use-after-free in fsl_mc_bus_remove()
  mei: me: add raptor lake point S DID
  mei: hbm: drop capability response on early shutdown
  mei: me: set internal pg flag to off on hardware reset
  misc: rtsx: Fix clang -Wsometimes-uninitialized in rts5261_init_from_hw()
  comedi: vmk80xx: fix expression for tx buffer size

2 years agoMerge tag 'i2c-for-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sun, 19 Jun 2022 14:35:09 +0000 (09:35 -0500)]
Merge tag 'i2c-for-5.19-rc3' of git://git./linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "MAINTAINERS rectifications and a few minor driver fixes"

* tag 'i2c-for-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: mediatek: Fix an error handling path in mtk_i2c_probe()
  i2c: designware: Use standard optional ref clock implementation
  MAINTAINERS: core DT include belongs to core
  MAINTAINERS: add include/dt-bindings/i2c to I2C SUBSYSTEM HOST DRIVERS
  i2c: npcm7xx: Add check for platform_driver_register
  MAINTAINERS: Update Synopsys DesignWare I2C to Supported

2 years agoMerge tag 'xfs-5.19-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Sun, 19 Jun 2022 14:24:49 +0000 (09:24 -0500)]
Merge tag 'xfs-5.19-fixes-1' of git://git./fs/xfs/xfs-linux

Pull xfs fixes from Darrick Wong:
 "There's not a whole lot this time around (I'm still on vacation) but
  here are some important fixes for new features merged in -rc1:

   - Fix a bug where inode flag changes would accidentally drop nrext64

   - Fix a race condition when toggling LARP mode"

* tag 'xfs-5.19-fixes-1' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: preserve DIFLAG2_NREXT64 when setting other inode attributes
  xfs: fix variable state usage
  xfs: fix TOCTOU race involving the new logged xattrs control knob

2 years agoMerge tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 19 Jun 2022 02:51:12 +0000 (21:51 -0500)]
Merge tag 'ext4_for_linus_stable' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 fixes from Ted Ts'o:
 "Fix a variety of bugs, many of which were found by folks using fuzzing
  or error injection.

  Also fix up how test_dummy_encryption mount option is handled for the
  new mount API.

  Finally, fix/cleanup a number of comments and ext4 Documentation
  files"

* tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: fix a doubled word "need" in a comment
  ext4: add reserved GDT blocks check
  ext4: make variable "count" signed
  ext4: correct the judgment of BUG in ext4_mb_normalize_request
  ext4: fix bug_on ext4_mb_use_inode_pa
  ext4: fix up test_dummy_encryption handling for new mount API
  ext4: use kmemdup() to replace kmalloc + memcpy
  ext4: fix super block checksum incorrect after mount
  ext4: improve write performance with disabled delalloc
  ext4: fix warning when submitting superblock in ext4_commit_super()
  ext4, doc: remove unnecessary escaping
  ext4: fix incorrect comment in ext4_bio_write_page()
  fs: fix jbd2_journal_try_to_free_buffers() kernel-doc comment

2 years agoMerge tag '5.19-rc2-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sun, 19 Jun 2022 02:44:44 +0000 (21:44 -0500)]
Merge tag '5.19-rc2-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs client fixes from Steve French:
 "Two cifs debugging improvements - one found to deal with debugging a
  multichannel problem and one for a recent fallocate issue

  This does include the two larger multichannel reconnect (dynamically
  adjusting interfaces on reconnect) patches, because we recently found
  an additional problem with multichannel to one server type that I want
  to include at the same time"

* tag '5.19-rc2-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: when a channel is not found for server, log its connection id
  smb3: add trace point for SMB2_set_eof

2 years agoext4: fix a doubled word "need" in a comment
Xiang wangx [Sun, 5 Jun 2022 09:15:03 +0000 (17:15 +0800)]
ext4: fix a doubled word "need" in a comment

Signed-off-by: Xiang wangx <wangxiang@cdjrlc.com>
Link: https://lore.kernel.org/r/20220605091503.12513-1-wangxiang@cdjrlc.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 years agoext4: add reserved GDT blocks check
Zhang Yi [Wed, 1 Jun 2022 09:27:17 +0000 (17:27 +0800)]
ext4: add reserved GDT blocks check

We capture a NULL pointer issue when resizing a corrupt ext4 image which
is freshly clear resize_inode feature (not run e2fsck). It could be
simply reproduced by following steps. The problem is because of the
resize_inode feature was cleared, and it will convert the filesystem to
meta_bg mode in ext4_resize_fs(), but the es->s_reserved_gdt_blocks was
not reduced to zero, so could we mistakenly call reserve_backup_gdb()
and passing an uninitialized resize_inode to it when adding new group
descriptors.

 mkfs.ext4 /dev/sda 3G
 tune2fs -O ^resize_inode /dev/sda #forget to run requested e2fsck
 mount /dev/sda /mnt
 resize2fs /dev/sda 8G

 ========
 BUG: kernel NULL pointer dereference, address: 0000000000000028
 CPU: 19 PID: 3243 Comm: resize2fs Not tainted 5.18.0-rc7-00001-gfde086c5ebfd #748
 ...
 RIP: 0010:ext4_flex_group_add+0xe08/0x2570
 ...
 Call Trace:
  <TASK>
  ext4_resize_fs+0xbec/0x1660
  __ext4_ioctl+0x1749/0x24e0
  ext4_ioctl+0x12/0x20
  __x64_sys_ioctl+0xa6/0x110
  do_syscall_64+0x3b/0x90
  entry_SYSCALL_64_after_hwframe+0x44/0xae
 RIP: 0033:0x7f2dd739617b
 ========

The fix is simple, add a check in ext4_resize_begin() to make sure that
the es->s_reserved_gdt_blocks is zero when the resize_inode feature is
disabled.

Cc: stable@kernel.org
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Ritesh Harjani <ritesh.list@gmail.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220601092717.763694-1-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 years agoext4: make variable "count" signed
Ding Xiang [Mon, 30 May 2022 10:00:47 +0000 (18:00 +0800)]
ext4: make variable "count" signed

Since dx_make_map() may return -EFSCORRUPTED now, so change "count" to
be a signed integer so we can correctly check for an error code returned
by dx_make_map().

Fixes: 46c116b920eb ("ext4: verify dir block before splitting it")
Cc: stable@kernel.org
Signed-off-by: Ding Xiang <dingxiang@cmss.chinamobile.com>
Link: https://lore.kernel.org/r/20220530100047.537598-1-dingxiang@cmss.chinamobile.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 years agoext4: correct the judgment of BUG in ext4_mb_normalize_request
Baokun Li [Sat, 28 May 2022 11:00:16 +0000 (19:00 +0800)]
ext4: correct the judgment of BUG in ext4_mb_normalize_request

ext4_mb_normalize_request() can move logical start of allocated blocks
to reduce fragmentation and better utilize preallocation. However logical
block requested as a start of allocation (ac->ac_o_ex.fe_logical) should
always be covered by allocated blocks so we should check that by
modifying and to or in the assertion.

Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Ritesh Harjani <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20220528110017.354175-3-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 years agoext4: fix bug_on ext4_mb_use_inode_pa
Baokun Li [Sat, 28 May 2022 11:00:15 +0000 (19:00 +0800)]
ext4: fix bug_on ext4_mb_use_inode_pa

Hulk Robot reported a BUG_ON:
==================================================================
kernel BUG at fs/ext4/mballoc.c:3211!
[...]
RIP: 0010:ext4_mb_mark_diskspace_used.cold+0x85/0x136f
[...]
Call Trace:
 ext4_mb_new_blocks+0x9df/0x5d30
 ext4_ext_map_blocks+0x1803/0x4d80
 ext4_map_blocks+0x3a4/0x1a10
 ext4_writepages+0x126d/0x2c30
 do_writepages+0x7f/0x1b0
 __filemap_fdatawrite_range+0x285/0x3b0
 file_write_and_wait_range+0xb1/0x140
 ext4_sync_file+0x1aa/0xca0
 vfs_fsync_range+0xfb/0x260
 do_fsync+0x48/0xa0
[...]
==================================================================

Above issue may happen as follows:
-------------------------------------
do_fsync
 vfs_fsync_range
  ext4_sync_file
   file_write_and_wait_range
    __filemap_fdatawrite_range
     do_writepages
      ext4_writepages
       mpage_map_and_submit_extent
        mpage_map_one_extent
         ext4_map_blocks
          ext4_mb_new_blocks
           ext4_mb_normalize_request
            >>> start + size <= ac->ac_o_ex.fe_logical
           ext4_mb_regular_allocator
            ext4_mb_simple_scan_group
             ext4_mb_use_best_found
              ext4_mb_new_preallocation
               ext4_mb_new_inode_pa
                ext4_mb_use_inode_pa
                 >>> set ac->ac_b_ex.fe_len <= 0
           ext4_mb_mark_diskspace_used
            >>> BUG_ON(ac->ac_b_ex.fe_len <= 0);

we can easily reproduce this problem with the following commands:
`fallocate -l100M disk`
`mkfs.ext4 -b 1024 -g 256 disk`
`mount disk /mnt`
`fsstress -d /mnt -l 0 -n 1000 -p 1`

The size must be smaller than or equal to EXT4_BLOCKS_PER_GROUP.
Therefore, "start + size <= ac->ac_o_ex.fe_logical" may occur
when the size is truncated. So start should be the start position of
the group where ac_o_ex.fe_logical is located after alignment.
In addition, when the value of fe_logical or EXT4_BLOCKS_PER_GROUP
is very large, the value calculated by start_off is more accurate.

Cc: stable@kernel.org
Fixes: cd648b8a8fd5 ("ext4: trim allocation requests to group size")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Ritesh Harjani <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20220528110017.354175-2-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 years agoext4: fix up test_dummy_encryption handling for new mount API
Eric Biggers [Thu, 26 May 2022 04:04:12 +0000 (21:04 -0700)]
ext4: fix up test_dummy_encryption handling for new mount API

Since ext4 was converted to the new mount API, the test_dummy_encryption
mount option isn't being handled entirely correctly, because the needed
fscrypt_set_test_dummy_encryption() helper function combines
parsing/checking/applying into one function.  That doesn't work well
with the new mount API, which split these into separate steps.

This was sort of okay anyway, due to the parsing logic that was copied
from fscrypt_set_test_dummy_encryption() into ext4_parse_param(),
combined with an additional check in ext4_check_test_dummy_encryption().
However, these overlooked the case of changing the value of
test_dummy_encryption on remount, which isn't allowed but ext4 wasn't
detecting until ext4_apply_options() when it's too late to fail.
Another bug is that if test_dummy_encryption was specified multiple
times with an argument, memory was leaked.

Fix this up properly by using the new helper functions that allow
splitting up the parse/check/apply steps for test_dummy_encryption.

Fixes: cebe85d570cf ("ext4: switch to the new mount api")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Link: https://lore.kernel.org/r/20220526040412.173025-1-ebiggers@kernel.org
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 years agoext4: use kmemdup() to replace kmalloc + memcpy
Shuqi Zhang [Wed, 25 May 2022 03:01:20 +0000 (11:01 +0800)]
ext4: use kmemdup() to replace kmalloc + memcpy

Replace kmalloc + memcpy with kmemdup()

Signed-off-by: Shuqi Zhang <zhangshuqi3@huawei.com>
Reviewed-by: Ritesh Harjani <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20220525030120.803330-1-zhangshuqi3@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 years agoext4: fix super block checksum incorrect after mount
Ye Bin [Wed, 25 May 2022 01:29:04 +0000 (09:29 +0800)]
ext4: fix super block checksum incorrect after mount

We got issue as follows:
[home]# mount  /dev/sda  test
EXT4-fs (sda): warning: mounting fs with errors, running e2fsck is recommended
[home]# dmesg
EXT4-fs (sda): warning: mounting fs with errors, running e2fsck is recommended
EXT4-fs (sda): Errors on filesystem, clearing orphan list.
EXT4-fs (sda): recovery complete
EXT4-fs (sda): mounted filesystem with ordered data mode. Quota mode: none.
[home]# debugfs /dev/sda
debugfs 1.46.5 (30-Dec-2021)
Checksum errors in superblock!  Retrying...

Reason is ext4_orphan_cleanup will reset ‘s_last_orphan’ but not update
super block checksum.

To solve above issue, defer update super block checksum after
ext4_orphan_cleanup.

Signed-off-by: Ye Bin <yebin10@huawei.com>
Cc: stable@kernel.org
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Ritesh Harjani <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20220525012904.1604737-1-yebin10@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2 years agocifs: when a channel is not found for server, log its connection id
Shyam Prasad N [Tue, 14 Jun 2022 11:47:24 +0000 (11:47 +0000)]
cifs: when a channel is not found for server, log its connection id

cifs_ses_get_chan_index gets the index for a given server pointer.
When a match is not found, we warn about a possible bug.
However, printing details about the non-matching server could be
more useful to debug here.

Signed-off-by: Shyam Prasad N <sprasad@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
2 years agox86/tdx: Handle load_unaligned_zeropad() page-cross to a shared page
Kirill A. Shutemov [Tue, 14 Jun 2022 12:01:35 +0000 (15:01 +0300)]
x86/tdx: Handle load_unaligned_zeropad() page-cross to a shared page

load_unaligned_zeropad() can lead to unwanted loads across page boundaries.
The unwanted loads are typically harmless. But, they might be made to
totally unrelated or even unmapped memory. load_unaligned_zeropad()
relies on exception fixup (#PF, #GP and now #VE) to recover from these
unwanted loads.

In TDX guests, the second page can be shared page and a VMM may configure
it to trigger #VE.

The kernel assumes that #VE on a shared page is an MMIO access and tries to
decode instruction to handle it. In case of load_unaligned_zeropad() it
may result in confusion as it is not MMIO access.

Fix it by detecting split page MMIO accesses and failing them.
load_unaligned_zeropad() will recover using exception fixups.

The issue was discovered by analysis and reproduced artificially. It was
not triggered during testing.

[ dhansen: fix up changelogs and comments for grammar and clarity,
   plus incorporate Kirill's off-by-one fix]

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Link: https://lkml.kernel.org/r/20220614120135.14812-4-kirill.shutemov@linux.intel.com
2 years agoMerge tag 'nfs-for-5.19-2' of git://git.linux-nfs.org/projects/anna/linux-nfs
Linus Torvalds [Fri, 17 Jun 2022 20:17:57 +0000 (15:17 -0500)]
Merge tag 'nfs-for-5.19-2' of git://git.linux-nfs.org/projects/anna/linux-nfs

Pull NFS client fixes from Anna Schumaker:

 - Add FMODE_CAN_ODIRECT support to NFSv4 so opens don't fail

 - Fix trunking detection & cl_max_connect setting

 - Avoid pnfs_update_layout() livelocks

 - Don't keep retrying pNFS if the server replies with NFS4ERR_UNAVAILABLE

* tag 'nfs-for-5.19-2' of git://git.linux-nfs.org/projects/anna/linux-nfs:
  NFSv4: Add FMODE_CAN_ODIRECT after successful open of a NFS4.x file
  sunrpc: set cl_max_connect when cloning an rpc_clnt
  pNFS: Avoid a live lock condition in pnfs_update_layout()
  pNFS: Don't keep retrying if the server replied NFS4ERR_LAYOUTUNAVAILABLE

2 years agoMerge tag 'pci-v5.19-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
Linus Torvalds [Fri, 17 Jun 2022 20:12:20 +0000 (15:12 -0500)]
Merge tag 'pci-v5.19-fixes-2' of git://git./linux/kernel/git/helgaas/pci

Pull pci fix from Bjorn Helgaas:
 "Revert clipping of PCI host bridge windows to avoid E820 regions,
  which broke several machines by forcing unnecessary BAR reassignments
  (Hans de Goede)"

* tag 'pci-v5.19-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  x86/PCI: Revert "x86/PCI: Clip only host bridge windows for E820 regions"

2 years agoMerge tag 'printk-for-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 17 Jun 2022 19:57:42 +0000 (14:57 -0500)]
Merge tag 'printk-for-5.19-rc3' of git://git./linux/kernel/git/printk/linux

Pull printk fixes from Petr Mladek:
 "Make the global console_sem available for CPU that is handling panic()
  or shutdown.

  This is an old problem when an existing console lock owner might block
  console output, but it became more visible with the kthreads"

* tag 'printk-for-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/printk/linux:
  printk: Wait for the global console lock when the system is going down
  printk: Block console kthreads when direct printing will be required

2 years agox86/PCI: Revert "x86/PCI: Clip only host bridge windows for E820 regions"
Hans de Goede [Sun, 12 Jun 2022 14:43:25 +0000 (16:43 +0200)]
x86/PCI: Revert "x86/PCI: Clip only host bridge windows for E820 regions"

This reverts commit 4c5e242d3e93.

Prior to 4c5e242d3e93 ("x86/PCI: Clip only host bridge windows for E820
regions"), E820 regions did not affect PCI host bridge windows.  We only
looked at E820 regions and avoided them when allocating new MMIO space.
If firmware PCI bridge window and BAR assignments used E820 regions, we
left them alone.

After 4c5e242d3e93, we removed E820 regions from the PCI host bridge
windows before looking at BARs, so firmware assignments in E820 regions
looked like errors, and we moved things around to fit in the space left
(if any) after removing the E820 regions.  This unnecessary BAR
reassignment broke several machines.

Guilherme reported that Steam Deck fails to boot after 4c5e242d3e93.  We
clipped the window that contained most 32-bit BARs:

  BIOS-e820: [mem 0x00000000a0000000-0x00000000a00fffff] reserved
  acpi PNP0A08:00: clipped [mem 0x80000000-0xf7ffffff window] to [mem 0xa0100000-0xf7ffffff window] for e820 entry [mem 0xa0000000-0xa00fffff]

which forced us to reassign all those BARs, for example, this NVMe BAR:

  pci 0000:00:01.2: PCI bridge to [bus 01]
  pci 0000:00:01.2:   bridge window [mem 0x80600000-0x806fffff]
  pci 0000:01:00.0: BAR 0: [mem 0x80600000-0x80603fff 64bit]
  pci 0000:00:01.2: can't claim window [mem 0x80600000-0x806fffff]: no compatible bridge window
  pci 0000:01:00.0: can't claim BAR 0 [mem 0x80600000-0x80603fff 64bit]: no compatible bridge window

  pci 0000:00:01.2: bridge window: assigned [mem 0xa0100000-0xa01fffff]
  pci 0000:01:00.0: BAR 0: assigned [mem 0xa0100000-0xa0103fff 64bit]

All the reassignments were successful, so the devices should have been
functional at the new addresses, but some were not.

Andy reported a similar failure on an Intel MID platform.  Benjamin
reported a similar failure on a VMWare Fusion VM.

Note: this is not a clean revert; this revert keeps the later change to
make the clipping dependent on a new pci_use_e820 bool, moving the checking
of this bool to arch_remove_reservations().

[bhelgaas: commit log, add more reporters and testers]
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=216109
Reported-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
Reported-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Reported-by: Benjamin Coddington <bcodding@redhat.com>
Reported-by: Jongman Heo <jongman.heo@gmail.com>
Fixes: 4c5e242d3e93 ("x86/PCI: Clip only host bridge windows for E820 regions")
Link: https://lore.kernel.org/r/20220612144325.85366-1-hdegoede@redhat.com
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
Tested-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Tested-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 17 Jun 2022 18:55:19 +0000 (13:55 -0500)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Catalin Marinas:

 - Revert the moving of the jump labels initialisation before
   setup_machine_fdt(). The bug was fixed in drivers/char/random.c.

 - Ftrace fixes: branch range check and consistent handling of PLTs.

 - Clean rather than invalidate FROM_DEVICE buffers at start of DMA
   transfer (safer if such buffer is mapped in user space). A cache
   invalidation is done already at the end of the transfer.

 - A couple of clean-ups (unexport symbol, remove unused label).

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: mm: Don't invalidate FROM_DEVICE buffers at start of DMA transfer
  arm64/cpufeature: Unexport set_cpu_feature()
  arm64: ftrace: remove redundant label
  arm64: ftrace: consistently handle PLTs.
  arm64: ftrace: fix branch range checks
  Revert "arm64: Initialize jump labels before setup_machine_fdt()"

2 years agoMerge tag 'loongarch-fixes-5.19-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 17 Jun 2022 18:50:24 +0000 (13:50 -0500)]
Merge tag 'loongarch-fixes-5.19-2' of git://git./linux/kernel/git/chenhuacai/linux-loongson

Pull LoongArch fixes from Huacai Chen:
 "Add missing ELF_DETAILS in vmlinux.lds.S and fix document rendering"

* tag 'loongarch-fixes-5.19-2' of git://git.kernel.org/pub/scm/linux/kernel/git/chenhuacai/linux-loongson:
  docs/zh_CN/LoongArch: Fix notes rendering by using reST directives
  docs/LoongArch: Fix notes rendering by using reST directives
  LoongArch: vmlinux.lds.S: Add missing ELF_DETAILS

2 years agoMerge tag 'riscv-for-linus-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 17 Jun 2022 18:45:47 +0000 (13:45 -0500)]
Merge tag 'riscv-for-linus-5.19-rc3' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Palmer Dabbelt:

 - A fix for the PolarFire SOC's device tree

 - A handful of fixes for the recently added Svpmbt support

 - An improvement to the Kconfig text for Svpbmt

* tag 'riscv-for-linus-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: Improve description for RISCV_ISA_SVPBMT Kconfig symbol
  riscv: drop cpufeature_apply_feature tracking variable
  riscv: fix dependency for t-head errata
  riscv: dts: microchip: re-add pdma to mpfs device tree

2 years agoMerge tag 'hyperv-fixes-signed-20220617' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 17 Jun 2022 18:39:12 +0000 (13:39 -0500)]
Merge tag 'hyperv-fixes-signed-20220617' of git://git./linux/kernel/git/hyperv/linux

Pull hyperv fixes from Wei Liu:

 - Fix hv_init_clocksource annotation (Masahiro Yamada)

 - Two bug fixes for vmbus driver (Saurabh Sengar)

 - Fix SEV negotiation (Tianyu Lan)

 - Fix comments in code (Xiang Wang)

 - One minor fix to HID driver (Michael Kelley)

* tag 'hyperv-fixes-signed-20220617' of git://git.kernel.org/pub/scm/linux/kernel/git/hyperv/linux:
  x86/Hyper-V: Add SEV negotiate protocol support in Isolation VM
  Drivers: hv: vmbus: Release cpu lock in error case
  HID: hyperv: Correctly access fields declared as __le16
  clocksource: hyper-v: unexport __init-annotated hv_init_clocksource()
  Drivers: hv: Fix syntax errors in comments
  Drivers: hv: vmbus: Don't assign VMbus channel interrupts to isolated CPUs

2 years agoMerge tag 'block-5.19-2022-06-16' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 17 Jun 2022 18:22:58 +0000 (11:22 -0700)]
Merge tag 'block-5.19-2022-06-16' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:

 - NVMe pull request from Christoph
      - Quirks, quirks, quirks to work around buggy consumer grade
        devices (Keith Bush, Ning Wang, Stefan Reiter, Rasheed Hsueh)
      - Better kernel messages for devices that need quirking (Keith
        Bush)
      - Make a kernel message more useful (Thomas Weißschuh)

 - MD pull request from Song, with a few fixes

 - blk-mq sysfs locking fixes (Ming)

 - BFQ stats fix (Bart)

 - blk-mq offline queue fix (Bart)

 - blk-mq flush request tag fix (Ming)

* tag 'block-5.19-2022-06-16' of git://git.kernel.dk/linux-block:
  block/bfq: Enable I/O statistics
  blk-mq: don't clear flush_rq from tags->rqs[]
  blk-mq: avoid to touch q->elevator without any protection
  blk-mq: protect q->elevator by ->sysfs_lock in blk_mq_elv_switch_none
  block: Fix handling of offline queues in blk_mq_alloc_request_hctx()
  md/raid5-ppl: Fix argument order in bio_alloc_bioset()
  Revert "md: don't unregister sync_thread with reconfig_mutex held"
  nvme-pci: disable write zeros support on UMIC and Samsung SSDs
  nvme-pci: avoid the deepest sleep state on ZHITAI TiPro7000 SSDs
  nvme-pci: sk hynix p31 has bogus namespace ids
  nvme-pci: smi has bogus namespace ids
  nvme-pci: phison e12 has bogus namespace ids
  nvme-pci: add NVME_QUIRK_BOGUS_NID for ADATA XPG GAMMIX S50
  nvme-pci: add trouble shooting steps for timeouts
  nvme: add bug report info for global duplicate id
  nvme: add device name to warning in uuid_show()

2 years agoMerge tag 'io_uring-5.19-2022-06-16' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 17 Jun 2022 18:14:07 +0000 (11:14 -0700)]
Merge tag 'io_uring-5.19-2022-06-16' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:
 "Bigger than usual at this time, both because we missed -rc2, but also
  because of some reverts that we chose to do. In detail:

   - Adjust mapped buffer API while we still can (Dylan)

   - Mapped buffer fixes (Dylan, Hao, Pavel, me)

   - Fix for uring_cmd wrong API usage for task_work (Dylan)

   - Fix for bug introduced in fixed file closing (Hao)

   - Fix race in buffer/file resource handling (Pavel)

   - Revert the NOP support for CQE32 and buffer selection that was
     brought up during the merge window (Pavel)

   - Remove IORING_CLOSE_FD_AND_FILE_SLOT introduced in this merge
     window. The API needs further refining, so just yank it for now and
     we'll revisit for a later kernel.

   - Series cleaning up the CQE32 support added in this merge window,
     making it more integrated rather than sitting on the side (Pavel)"

* tag 'io_uring-5.19-2022-06-16' of git://git.kernel.dk/linux-block: (21 commits)
  io_uring: recycle provided buffer if we punt to io-wq
  io_uring: do not use prio task_work_add in uring_cmd
  io_uring: commit non-pollable provided mapped buffers upfront
  io_uring: make io_fill_cqe_aux honour CQE32
  io_uring: remove __io_fill_cqe() helper
  io_uring: fix ->extra{1,2} misuse
  io_uring: fill extra big cqe fields from req
  io_uring: unite fill_cqe and the 32B version
  io_uring: get rid of __io_fill_cqe{32}_req()
  io_uring: remove IORING_CLOSE_FD_AND_FILE_SLOT
  Revert "io_uring: add buffer selection support to IORING_OP_NOP"
  Revert "io_uring: support CQE32 for nop operation"
  io_uring: limit size of provided buffer ring
  io_uring: fix types in provided buffer ring
  io_uring: fix index calculation
  io_uring: fix double unlock for pbuf select
  io_uring: kbuf: fix bug of not consuming ring buffer in partial io case
  io_uring: openclose: fix bug of closing wrong fixed file
  io_uring: fix not locked access to fixed buf table
  io_uring: fix races with buffer table unregister
  ...

2 years agoarm64: mm: Don't invalidate FROM_DEVICE buffers at start of DMA transfer
Will Deacon [Fri, 10 Jun 2022 15:12:27 +0000 (16:12 +0100)]
arm64: mm: Don't invalidate FROM_DEVICE buffers at start of DMA transfer

Invalidating the buffer memory in arch_sync_dma_for_device() for
FROM_DEVICE transfers

When using the streaming DMA API to map a buffer prior to inbound
non-coherent DMA (i.e. DMA_FROM_DEVICE), we invalidate any dirty CPU
cachelines so that they will not be written back during the transfer and
corrupt the buffer contents written by the DMA. This, however, poses two
potential problems:

  (1) If the DMA transfer does not write to every byte in the buffer,
      then the unwritten bytes will contain stale data once the transfer
      has completed.

  (2) If the buffer has a virtual alias in userspace, then stale data
      may be visible via this alias during the period between performing
      the cache invalidation and the DMA writes landing in memory.

Address both of these issues by cleaning (aka writing-back) the dirty
lines in arch_sync_dma_for_device(DMA_FROM_DEVICE) instead of discarding
them using invalidation.

Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Russell King <linux@armlinux.org.uk>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220606152150.GA31568@willie-the-truck
Signed-off-by: Will Deacon <will@kernel.org>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20220610151228.4562-2-will@kernel.org
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
2 years agoMerge tag 'fs_for_v5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/jack...
Linus Torvalds [Fri, 17 Jun 2022 17:09:24 +0000 (10:09 -0700)]
Merge tag 'fs_for_v5.19-rc3' of git://git./linux/kernel/git/jack/linux-fs

Pull writeback and ext2 fixes from Jan Kara:
 "A fix for writeback bug which prevented machines with kdevtmpfs from
  booting and also one small ext2 bugfix in IO error handling"

* tag 'fs_for_v5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  init: Initialize noop_backing_dev_info early
  ext2: fix fs corruption when trying to remove a non-empty directory with IO error

2 years agoMerge tag 'for-5.19/dm-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 17 Jun 2022 17:03:53 +0000 (10:03 -0700)]
Merge tag 'for-5.19/dm-fixes-3' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:

 - Fix a race in DM core's dm_start_io_acct that could result in double
   accounting for abnormal IO (e.g. discards, write zeroes, etc).

 - Fix a use-after-free in DM core's dm_put_live_table_bio.

 - Fix a race for REQ_NOWAIT bios being issued despite no support from
   underlying DM targets (due to DM table reload at an "unlucky" time)

 - Fix access beyond allocated bitmap in DM mirror's log.

* tag 'for-5.19/dm-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm mirror log: round up region bitmap size to BITS_PER_LONG
  dm: fix narrow race for REQ_NOWAIT bios being issued despite no support
  dm: fix use-after-free in dm_put_live_table_bio
  dm: fix race in dm_start_io_acct

2 years agoMerge tag 'hwmon-for-v5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 17 Jun 2022 17:02:26 +0000 (10:02 -0700)]
Merge tag 'hwmon-for-v5.19-rc3' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon fixes from Guenter Roeck:

 - Add missing lock protection in occ driver

 - Add missing comma in board name list in asus-ec-sensors driver

 - Fix devicetree bindings for ti,tmp401

* tag 'hwmon-for-v5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (asus-ec-sensors) add missing comma in board name list.
  hwmon: (occ) Lock mutex in shutdown to prevent race with occ_active
  dt-bindings: hwmon: ti,tmp401: Drop 'items' from 'ti,n-factor' property

2 years agoMerge tag 'linux-watchdog-5.19-rc3' of git://www.linux-watchdog.org/linux-watchdog
Linus Torvalds [Fri, 17 Jun 2022 17:00:25 +0000 (10:00 -0700)]
Merge tag 'linux-watchdog-5.19-rc3' of git://linux-watchdog.org/linux-watchdog

Pull watchdog fix from Wim Van Sebroeck:
 "Add missing MODULE_LICENSE in gxp driver"

* tag 'linux-watchdog-5.19-rc3' of git://www.linux-watchdog.org/linux-watchdog:
  watchdog: gxp: Add missing MODULE_LICENSE

2 years agoMerge tag 'v5.19-p2' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Fri, 17 Jun 2022 15:27:27 +0000 (08:27 -0700)]
Merge tag 'v5.19-p2' of git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto fix from Herbert Xu:
 "This fixes a potential build failure when CRYPTO=m"

* tag 'v5.19-p2' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: memneq - move into lib/

2 years agoMerge tag 'char-misc-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregk...
Linus Torvalds [Fri, 17 Jun 2022 14:58:39 +0000 (07:58 -0700)]
Merge tag 'char-misc-5.19-rc3' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are some small char/misc driver fixes for 5.19-rc3 that resolve
  some reported issues.

  They include:

   - mei driver fixes

   - comedi driver fix

   - rtsx build warning fix

   - fsl-mc-bus driver fix

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'char-misc-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  eeprom: at25: Split reads into chunks and cap write size
  misc: atmel-ssc: Fix IRQ check in ssc_probe
  char: lp: remove redundant initialization of err

2 years agoMerge tag 'staging-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Fri, 17 Jun 2022 14:55:24 +0000 (07:55 -0700)]
Merge tag 'staging-5.19-rc3' of git://git./linux/kernel/git/gregkh/staging

Pull staging driver fixes from Greg KH:
 "Here are some small staging driver fixes for 5.19-rc3 that resolve
  reported issues:

   - remove visorbus.h which was forgotten in the -rc1 merge where the
     code that used it was removed

   - olpc_dcon: mark as broken to allow the DRM developers to evolve the
     fbdev api properly without having to deal with this obsolete
     driver. It will be removed soon if no one steps up to adopt it and
     fix the issues with it.

   - rtl8723bs driver fix

   - r8188eu driver fix to resolve many reports of the driver being
     broken with -rc1.

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'staging-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  staging: Also remove the Unisys visorbus.h
  staging: rtl8723bs: Allocate full pwep structure
  staging: olpc_dcon: mark driver as broken
  staging: r8188eu: Fix warning of array overflow in ioctl_linux.c
  staging: r8188eu: fix rtw_alloc_hwxmits error detection for now

2 years agoMerge tag 'tty-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Fri, 17 Jun 2022 14:52:43 +0000 (07:52 -0700)]
Merge tag 'tty-5.19-rc3' of git://git./linux/kernel/git/gregkh/tty

Pull tty/serial driver fixes from Greg KH:
 "Here are some small tty and serial driver fixes for 5.19-rc3 to
  resolve some reported problems:

   - 8250 lsr read bugfix

   - n_gsm line discipline allocation fix

   - qcom serial driver fix for reported lockups that happened in -rc1

   - goldfish tty driver fix

  All have been in linux-next for a while now with no reported issues"

* tag 'tty-5.19-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  serial: 8250: Store to lsr_save_flags after lsr read
  tty: goldfish: Fix free_irq() on remove
  tty: serial: qcom-geni-serial: Implement start_rx callback
  serial: core: Introduce callback for start_rx and do stop_rx in suspend only if this callback implementation is present.
  tty: n_gsm: Debug output allocation must use GFP_ATOMIC