kernel/linux.git/drivers/cxl, branch v6.2

cxl/region: Fix passthrough-decoder detection

2023-02-07T19:04:30+00:00

A passthrough decoder is a decoder that maps only 1 target. It is a special case because it does not impose any constraints on the interleave-math as compared to a decoder with multiple targets. Extend the passthrough case to multi-target-capable decoders that only have one target selected. I.e. the current code was only considering passthrough *ports* which are only a subset of the potential passthrough decoder scenarios. Fixes: e4f6dfa9ef75 ("cxl/region: Fix 'distance' calculation with passthrough ports") Cc: Reviewed-by: Dave Jiang Link: https://lore.kernel.org/r/167564540422.847146.13816934143225777888.stgit@dwillia2-xfh.jf.intel.com Signed-off-by: Dan Williams

cxl/region: Fix null pointer dereference for resetting decoder

2023-02-07T01:33:50+00:00

Not all decoders have a reset callback. The CXL specification allows a host bridge with a single root port to have no explicit HDM decoders. Currently the region driver assumes there are none. As such the CXL core creates a special pass through decoder instance without a commit/reset callback. Prior to this patch, the ->reset() callback was called unconditionally when calling cxl_region_decode_reset. Thus a configuration with 1 Host Bridge, 1 Root Port, and one directly attached CXL type 3 device or multiple CXL type 3 devices attached to downstream ports of a switch can cause a null pointer dereference. Before the fix, a kernel crash was observed when we destroy the region, and a pass through decoder is reset. The issue can be reproduced as below, 1) create a region with a CXL setup which includes a HB with a single root port under which a memdev is attached directly. 2) destroy the region with cxl destroy-region regionX -f. Fixes: 176baefb2eb5 ("cxl/hdm: Commit decoder state to hardware") Cc: Signed-off-by: Fan Ni Reviewed-by: Davidlohr Bueso Reviewed-by: Dave Jiang Reviewed-by: Jonathan Cameron Tested-by: Gregory Price Reviewed-by: Gregory Price Link: https://lore.kernel.org/r/20221215170909.2650271-1-fan.ni@samsung.com Signed-off-by: Dan Williams

cxl/pmem: Fix nvdimm unregistration when cxl_pmem driver is absent

2023-01-25T23:35:26+00:00

The cxl_pmem.ko module houses the driver for both cxl_nvdimm_bridge objects and cxl_nvdimm objects. When the core creates a cxl_nvdimm it arranges for it to be autoremoved when the bridge goes down. However, if the bridge never initialized because the cxl_pmem.ko module never loaded, it sets up a the following crash scenario: BUG: kernel NULL pointer dereference, address: 0000000000000478 [..] RIP: 0010:cxl_nvdimm_probe+0x99/0x140 [cxl_pmem] [..] Call Trace: cxl_bus_probe+0x17/0x50 [cxl_core] really_probe+0xde/0x380 __driver_probe_device+0x78/0x170 driver_probe_device+0x1f/0x90 __driver_attach+0xd2/0x1c0 bus_for_each_dev+0x79/0xc0 bus_add_driver+0x1b1/0x200 driver_register+0x89/0xe0 cxl_pmem_init+0x50/0xff0 [cxl_pmem] It turns out the recent rework to simplify nvdimm probing obviated the need to unregister cxl_nvdimm objects at cxl_nvdimm_bridge ->remove() time. Leave the cxl_nvdimm device registered until the hosting cxl_memdev departs. The alternative is that the cxl_memdev needs to be reattached whenever the cxl_nvdimm_bridge attach state cycles, which is awkward and unnecessary. The only requirement is to make sure that when the cxl_nvdimm_bridge goes away any dependent cxl_nvdimm objects are shutdown. Handle that in unregister_nvdimm_bus(). With these registration entanglements removed there is no longer a need to pre-load the cxl_pmem module in cxl_acpi. Fixes: cb9cfff82f6a ("cxl/acpi: Simplify cxl_nvdimm_bridge probing") Reported-by: Gregory Price Debugged-by: Jonathan Cameron Tested-by: Jonathan Cameron Reviewed-by: Jonathan Cameron Reviewed-by: Dave Jiang Link: https://lore.kernel.org/r/167426077263.3955046.9695309346988027311.stgit@dwillia2-xfh.jf.intel.com Signed-off-by: Dan Williams

cxl: fix cxl_report_and_clear() RAS UE addr mis-assignment

2023-01-09T20:55:54+00:00

'addr' that contains RAS UE register address is re-assigned to RAS_CAP_CONTROL offset if there are multiple UE errors. Use different addr variable to avoid the reassignment mistake. Fixes: 2905cb5236cb ("cxl/pci: Add (hopeful) error handling support") Reported-by: Jonathan Cameron Signed-off-by: Dave Jiang Reviewed-by: Jonathan Cameron Reviewed-by: Ira Weiny Link: https://lore.kernel.org/r/167302318779.580155.15233596744650706167.stgit@djiang5-mobl3.local Signed-off-by: Dan Williams

cxl/region: Fix memdev reuse check

2022-12-08T21:03:47+00:00

Due to a typo, the check of whether or not a memdev has already been used as a target for the region (above code piece) will always be skipped. Given a memdev with more than one HDM decoder, an interleaved region can be created that maps multiple HPAs to the same DPA. According to CXL spec 3.0 8.1.3.8.4, "Aliasing (mapping more than one Host Physical Address (HPA) to a single Device Physical Address) is forbidden." Fix this by using existing iterator for memdev reuse check. Cc: Fixes: 384e624bb211 ("cxl/region: Attach endpoint decoders") Signed-off-by: Fan Ni Link: https://lore.kernel.org/r/20221107212153.745993-1-fan.ni@samsung.com Signed-off-by: Dan Williams

cxl/pci: Remove endian confusion

2022-12-06T22:38:12+00:00

readl() already handles endian conversion. That's the main difference between readl() and __raw_readl(). This is benign on little-endian systems, but big endian systems will end up byte-swabbing twice. Fixes: 2905cb5236cb ("cxl/pci: Add (hopeful) error handling support") Cc: Jonathan Cameron Cc: Dave Jiang Reviewed-by: Dave Jiang Reviewed-by: Jonathan Cameron Reviewed-by: Ira Weiny Link: https://lore.kernel.org/r/167030092025.4045167.10651070153523351093.stgit@dwillia2-xfh.jf.intel.com Signed-off-by: Dan Williams

cxl/pci: Add some type-safety to the AER trace points

2022-12-06T22:37:52+00:00

The first argument to the CXL AER trace points is the source device. Pass a 'const struct device *' rather than a 'const char *' for more type precision / safety. Cc: Jonathan Cameron Cc: Dave Jiang Cc: Steven Rostedt Reviewed-by: Dave Jiang Reviewed-by: Jonathan Cameron Reviewed-by: Ira Weiny Link: https://lore.kernel.org/r/167030091477.4045167.15174636482098463885.stgit@dwillia2-xfh.jf.intel.com Signed-off-by: Dan Williams

cxl/security: Drop security command ioctl uapi

2022-12-06T22:36:02+00:00

CXL PMEM security operations are routed through the NVDIMM sysfs interface. For this reason the corresponding commands are marked "exclusive" to preclude collisions between the ioctl ABI and the sysfs ABI. However, a better way to preclude that collision is to simply remove the ioctl ABI (command-id definitions) for those operations. Now that cxl_internal_send_cmd() (formerly cxl_mbox_send_cmd()) no longer needs to talk the cxl_mem_commands array, all of the uapi definitions for the security commands can be dropped. These never appeared in a released kernel, so no regression risk. Reviewed-by: Dave Jiang Reviewed-by: Ira Weiny Link: https://lore.kernel.org/r/167030056464.4044561.11486507095384253833.stgit@dwillia2-xfh.jf.intel.com Signed-off-by: Dan Williams

cxl/mbox: Add variable output size validation for internal commands

2022-12-06T22:36:02+00:00

cxl_internal_send_cmd() skips output size validation for variable output commands which is not ideal. Most of the time internal usages want to fail if the output size does not match what was requested. For other commands where the caller cannot predict the size there is usually a a header that conveys how much vaild data is in the payload. For those cases add @min_out as a parameter to specify what the minimum response payload needs to be for the caller to parse the rest of the payload. In this patch only Get Supported Logs has that behavior, but going forward records retrieval commands like Get Poison List and Get Event Records can use @min_out to retrieve a variable amount of records. Critically, this validation scheme skips the needs to interrogate the cxl_mem_commands array which in turn frees up the implementation to support internal command enabling without also enabling external / user commands. Reviewed-by: Dave Jiang Reviewed-by: Ira Weiny Link: https://lore.kernel.org/r/167030055918.4044561.10339573829837910505.stgit@dwillia2-xfh.jf.intel.com Signed-off-by: Dan Williams

cxl/mbox: Enable cxl_mbox_send_cmd() users to validate output size

2022-12-06T22:36:02+00:00

Internally cxl_mbox_send_cmd() converts all passed-in parameters to a 'struct cxl_mbox_cmd' instance and sends that to cxlds->mbox_send(). It then teases the possibilty that the caller can validate the output size. However, they cannot since the resulting output size is not conveyed to the called. Fix that by making the caller pass in a constructed 'struct cxl_mbox_cmd'. This prepares for a future patch to add output size validation on a per-command basis. Given the change in signature, also change the name to differentiate it from the user command submission path that performs more validation before generating the 'struct cxl_mbox_cmd' instance to execute. Reviewed-by: Dave Jiang Reviewed-by: Ira Weiny Link: https://lore.kernel.org/r/167030055370.4044561.17788093375112783036.stgit@dwillia2-xfh.jf.intel.com Signed-off-by: Dan Williams