- 23 Apr, 2023 20 commits
-
-
Dan Williams authored
Pick up late v6.3 fixes for v6.4.
-
Dan Williams authored
Include the poison list and injection infrastructure from Alison for v6.4.
-
Alison Schofield authored
The cxl_mem driver uses debugfs to support poison inject and clear. Add debugfs to the list of required symbols so that cxl_test can emulate those poison operations. Signed-off-by:
Alison Schofield <alison.schofield@intel.com> Reviewed-by:
Dave Jiang <dave.jiang@intel.com> Link: https://lore.kernel.org/r/4f3aab57fbf1cc3ccde2eb887c5d90566c8d0e90.1681874357.git.alison.schofield@intel.comSigned-off-by:
Dan Williams <dan.j.williams@intel.com>
-
Alison Schofield authored
CXL devices may report a maximum number of addresses that a device allows to be poisoned using poison injection. When cxl_test creates mock CXL memory devices, it defaults to MOCK_INJECT_DEV_MAX==88 for all mocked memdevs. Add a sysfs attribute, poison_inject_max to module cxl_mock_mem so that users can set a custom device injection limit. Fail, and return -EBUSY, if the mock poison list is not empty. /sys/bus/platform/drivers/cxl_mock_mem/poison_inject_max A simple usage model is to set the attribute before running a test in order to emulate a device's poison handling. Signed-off-by:
-
Alison Schofield authored
Prior to poison inject support, the mock of 'Get Poison List' returned a poison list containing a single mocked error record. Following the addition of poison inject and clear support to the mock driver, use the mock_poison_list[], rather than faking an error record. Mock_poison_list[] list tracks the actual poison inject and clear requests issued by userspace. Signed-off-by:
-
Alison Schofield authored
Mock the clear of poison by deleting the device:address entry from the mock_poison_list[]. Behave like a real CXL device and do not fail if the address is not in the poison list, but offer a dev_dbg() message. Signed-off-by:
Jonathan Cameron <Jonathan.Cameron@huawei.com> Reviewed-by:
-
Alison Schofield authored
Mock the injection of poison by storing the device:address entries in mock_poison_list[]. Enforce a limit of 8 poison injections per memdev device and 128 total entries for the cxl_test mock driver. Introducing the mock_poison[] list here, makes it available for use in the mock of Clear Poison, and the mock of Get Poison List. Signed-off-by:
-
Alison Schofield authored
Inject and Clear Poison commands are optionally supported by CXL memdev devices and are intended for use in debug environments only. Add debugfs attributes for user access. Documentation/ABI/testing/debugfs-cxl describes the usage. Signed-off-by:
-
Alison Schofield authored
The cxl_poison trace event allows users to view the history of poison list reads. With the addition of inject and clear poison capabilities, users will expect similar tracing. Add trace types 'Inject' and 'Clear' to the cxl_poison trace_event and trace successful operations only. If the driver finds that the DPA being injected or cleared of poison is mapped in a region, that region info is included in the cxl_poison trace event. Region reconfigurations can make this extra info useless if the debug operations are not carefully managed. Signed-off-by:
-
Alison Schofield authored
Inject and clear poison capabilities and intended for debug usage only. In order to be useful in debug environments, the driver needs to allow inject and clear operations on DPAs mapped in regions. dev_warn_once() when either operation occurs. Signed-off-by:
-
Alison Schofield authored
CXL devices optionally support the CLEAR POISON mailbox command. Add memdev driver support for clearing poison. Per the CXL Specification (3.0 8.2.9.8.4.3), after receiving a valid clear poison request, the device removes the address from the device's Poison List and writes 0 (zero) for 64 bytes starting at address. If the device cannot clear poison from the address, it returns a permanent media error and -ENXIO is returned to the user. Additionally, and per the spec also, it is not an error to clear poison of an address that is not poisoned. If the address is not contained in the device's dpa resource, or is not 64 byte aligned, the driver returns -EINVAL without sending the command to the device. Poison clearing is intended for debug only and will be exposed to userspace through debugfs. Restrict compilation to CONFIG_DEBUG_FS. Implementation note: Although the CXL specification defines the clear command to accept 64 bytes of 'write-data', this implementation always uses zeroes as write-data. Signed-off-by:
-
Alison Schofield authored
CXL devices optionally support the INJECT POISON mailbox command. Add memdev driver support for the mailbox command. Per the CXL Specification (3.0 8.2.9.8.4.2), after receiving a valid inject poison request, the device will return poison when the address is accessed through the CXL.mem driver. Injecting poison adds the address to the device's Poison List and the error source is set to Injected. In addition, the device adds a poison creation event to its internal Informational Event log, updates the Event Status register, and if configured, interrupts the host. Also, per the CXL Specification, it is not an error to inject poison into an address that already has poison present and no error is returned from the device. If the address is not contained in the device's dpa resource, or is not 64 byte aligned, return -EINVAL without issuing the mbox command. Poison injection is intended for debug only and will be exposed to userspace through debugfs. Restrict compilation to CONFIG_DEBUG_FS. Signed-off-by:
-
Alison Schofield authored
Make mock memdevs support the Get Poison List mailbox command. Return a fake poison error record when the get poison list command is issued. This supports testing the kernel tracing and cxl list capabilities for media errors. Signed-off-by:
Ira Weiny <ira.weiny@intel.com> Reviewed-by:
-
Alison Schofield authored
When a cxl_poison trace event is reported for a region, the poisoned Device Physical Address (DPA) can be translated to a Host Physical Address (HPA) for consumption by user space. Translate and add the resulting HPA to the cxl_poison trace event. Follow the device decode logic as defined in the CXL Spec 3.0 Section 8.2.4.19.13. If no region currently maps the poison, assign ULLONG_MAX to the cxl_poison event hpa field. Signed-off-by:
-
Alison Schofield authored
User space may need to know which region, if any, maps the poison address(es) logged in a cxl_poison trace event. Since the mapping of DPAs (device physical addresses) to a region can change, the kernel must provide this information at the time the poison list is read. The event informs user space that at event <timestamp> this <region> mapped to this <DPA>, which is poisoned. The cxl_poison trace event is already wired up to log the region name and uuid if it receives param 'struct cxl_region'. In order to provide that cxl_region, add another method for gathering poison - by committed endpoint decoder mappings. This method is only available with CONFIG_CXL_REGION and is only used if a region actually maps the memdev where poison is being read. After the region driver reads the poison list for all the mapped resources, poison is read for any remaining unmapped resources. The default method remains: read the poison by memdev resource. Signed-off-by:
-
Alison Schofield authored
When a boolean 'true' is written to this attribute the memdev driver retrieves the poison list from the device. The list consists of addresses that are poisoned, or would result in poison if accessed, and the source of the poison. This attribute is only visible for devices supporting the capability. The retrieved errors are logged as kernel events when cxl_poison event tracing is enabled. Signed-off-by:
-
Alison Schofield authored
CXL devices may support the retrieval of a device poison list. Add a new trace event that the CXL subsystem may use to log the media-error records returned in the poison list. Log each media-error record as a cxl_poison trace event of type 'List'. Signed-off-by:
-
Alison Schofield authored
CXL devices maintain a list of locations that are poisoned or result in poison if the addresses are accessed by the host. Per the spec, (CXL 3.0 8.2.9.8.4.1), the device returns this Poison list as a set of Media Error Records that include the source of the error, the starting device physical address, and length. The length is the number of adjacent DPAs in the record and is in units of 64 bytes. Retrieve the poison list. Signed-off-by:
-
Alison Schofield authored
Driver reads of the poison list are synchronized to ensure that a reader does not get an incomplete list because their request overlapped (was interrupted or preceded by) another read request of the same DPA range. (CXL Spec 3.0 Section 8.2.9.8.4.1). The driver maintains state information to achieve this goal. To initialize the state, first recognize the poison commands in the CEL (Command Effects Log). If the device supports Get Poison List, allocate a single buffer for the poison list and protect it with a lock. Signed-off-by:
-
Alison Schofield authored
The Get, Inject, and Clear poison commands are not available for direct user access because they require kernel driver controls to perform safely. Further restrict access to these commands by requiring the selection of the debugfs attribute 'cxl_raw_allow_all' to enable in raw mode. Signed-off-by:
-
- 22 Apr, 2023 2 commits
-
-
Dan Williams authored
The CXL subsystem is adding formal mechanisms for managing device poison. Minimize the maintenance burden going forward, and maximize the investment in common tooling by deprecating direct user access to poison commands outside of CXL_MEM_RAW_COMMANDS debug scenarios. A new cxl_deprecated_commands[] list is created for querying which command ids defined in previous kernels are now deprecated. CXL Media and Poison Management commands, opcodes 0x43XX, defined in CXL 3.0 Spec, Table 8-93 are deprecated with one exception: Get Scan Media Capabilities. Keep Get Scan Media Capabilities as it simply provides information and has no impact on the device state. Effectively all of the commands defined in: commit 87815ee9 ("cxl/pci: Add media provisioning required commands") ...were defined prematurely and should have waited until the kernel implementation was decided. To my knowledge there are no shipping devices with poison support and no known tools that would regress with this change. Co-developed-by:
-
Dan Williams authored
Not all endpoint CXL ports are associated with PCI devices. The cxl_test infrastructure models 'struct cxl_port' instances hosted by platform devices. Teach read_cdat_data() to be careful about non-pci hosted cxl_memdev instances. Otherwise, cxl_test crashes with this signature: RIP: 0010:xas_start+0x6d/0x290 [..] Call Trace: <TASK> xas_load+0xa/0x50 xas_find+0x25b/0x2f0 xa_find+0x118/0x1d0 pci_find_doe_mailbox+0x51/0xc0 read_cdat_data+0x45/0x190 [cxl_core] cxl_port_probe+0x10a/0x1e0 [cxl_port] cxl_bus_probe+0x17/0x50 [cxl_core] Some other cleanups are included like removing the single-use @uport variable, and removing the indirection through 'struct cxl_dev_state' to lookup the device that registered the memdev and may be a pci device. Fixes: af0a6c35 ("cxl/pci: Use CDAT DOE mailbox created by PCI core") Reviewed-by:
Lukas Wunner <lukas@wunner.de> Reviewed-by:
-
- 18 Apr, 2023 16 commits
-
-
Lukas Wunner authored
Jonathan notes that cxl_cdat_get_length() and cxl_cdat_read_table() allocate 32 dwords for the DOE response even though it may be smaller. In the case of cxl_cdat_get_length(), only the second dword of the response is of interest (it contains the length). So reduce the allocation to 2 dwords and let DOE discard the remainder. In the case of cxl_cdat_read_table(), a correctly sized allocation for the full CDAT already exists. Let DOE write each table entry directly into that allocation. There's a snag in that the table entry is preceded by a Table Access Response Header (1 dword, CXL 3.0 table 8-14). Save the last dword of the previous table entry, let DOE overwrite it with the header of the next entry and restore it afterwards. The resulting CDAT is preceded by 4 unavoidable useless bytes. Increase the allocation size accordingly. The buffer overflow check in cxl_cdat_read_table() becomes unnecessary because the remaining bytes in the allocation are tracked in "length", which is passed to DOE and limits how many bytes it writes to the allocation. Additionally, cxl_cdat_read_table() bails out if the DOE response is truncated due to insufficient space. Tested-by:
-
Dave Jiang authored
The cdat.table and cdat.length fields in struct cxl_port are set before CDAT retrieval and must therefore be unset on failure. Simplify by setting only on success. Suggested-by:
Bjorn Helgaas <bhelgaas@google.com> Link: https://lore.kernel.org/r/7a5c7104fb6a3016dbaec1c5d0ed34619ea11a0c.1678543498.git.lukas@wunner.deSigned-off-by:
-
Lukas Wunner authored
An upcoming user of DOE is CMA (Component Measurement and Authentication, PCIe r6.0 sec 6.31). It builds on SPDM (Security Protocol and Data Model): https://www.dmtf.org/dsp/DSP0274 SPDM message sizes are not always a multiple of dwords. To transport them over DOE without using bounce buffers, allow sending requests and receiving responses whose final dword is only partially populated. To be clear, PCIe r6.0 sec 6.30.1 specifies the Data Object Header 2 "Length" in dwords and pci_doe_send_req() and pci_doe_recv_resp() read/write dwords. So from a spec point of view, DOE is still specified in dwords and allowing non-dword request/response buffers is merely for the convenience of callers. Tested-by:
Ming Li <ming4.li@intel.com> Reviewed-by:
-
Lukas Wunner authored
The PCI core has just been amended to create a pci_doe_mb struct for every DOE instance on device enumeration. CXL (the only in-tree DOE user so far) has been migrated to use those mailboxes instead of creating its own. That leaves pcim_doe_create_mb() and pci_doe_for_each_off() without any callers, so drop them. pci_doe_supports_prot() is now only used internally, so declare it static. pci_doe_destroy_mb() is no longer used as callback for devm_add_action(), so refactor it to accept a struct pci_doe_mb pointer instead of a generic void pointer. Because pci_doe_create_mb() is only called on device enumeration, i.e. before driver binding, the workqueue name never contains a driver name. So replace dev_driver_string() with dev_bus_name() when generating the workqueue name. Tested-by:
-
Lukas Wunner authored
The PCI core has just been amended to create a pci_doe_mb struct for every DOE instance on device enumeration. Drop creation of a (duplicate) CDAT DOE mailbox on cxl probing in favor of the one already created by the PCI core. Tested-by:
-
Lukas Wunner authored
Currently a DOE instance cannot be shared by multiple drivers because each driver creates its own pci_doe_mb struct for a given DOE instance. For the same reason a DOE instance cannot be shared between the PCI core and a driver. Moreover, finding out which protocols a DOE instance supports requires creating a pci_doe_mb for it. If a device has multiple DOE instances, a driver looking for a specific protocol may need to create a pci_doe_mb for each of the device's DOE instances and then destroy those which do not support the desired protocol. That's obviously an inefficient way to do things. Overcome these issues by creating mailboxes in the PCI core on device enumeration. Provide a pci_find_doe_mailbox() API call to allow drivers to get a pci_doe_mb for a given (pci_dev, vendor, protocol) triple. This API is modeled after pci_find_capability() and can later be amended with a pci_find_next_doe_mailbox() call to iterate over all mailboxes of a given pci_dev which support a specific protocol. On removal, destroy the mailboxes in pci_destroy_dev(), after the driver is unbound. This allows drivers to use DOE in their ->remove() hook. On surprise removal, cancel ongoing DOE exchanges and prevent new ones from being scheduled. Thereby ensure that a hot-removed device doesn't needlessly wait for a running exchange to time out. Tested-by:
-
Lukas Wunner authored
DOE mailbox creation is currently only possible through a devres-managed API. The lifetime of mailboxes thus ends with driver unbinding. An upcoming commit will create DOE mailboxes upon device enumeration by the PCI core. Their lifetime shall not be limited by a driver. Therefore rework pcim_doe_create_mb() into the non-devres-managed pci_doe_create_mb(). Add pci_doe_destroy_mb() for mailbox destruction on device removal. Provide a devres-managed wrapper under the existing pcim_doe_create_mb() name. The error path of pcim_doe_create_mb() previously called xa_destroy() if alloc_ordered_workqueue() failed. That's unnecessary because the xarray is still empty at that point. It doesn't need to be destroyed until it's been populated by pci_doe_cache_protocols(). Arrange the error path of the new pci_doe_create_mb() accordingly. pci_doe_cancel_tasks() is no longer used as callback for devm_add_action(), so refactor it to accept a struct pci_doe_mb pointer instead of a generic void pointer. Tested-by:
-
Lukas Wunner authored
When a DOE mailbox is torn down, its workqueue is flushed once in pci_doe_flush_mb() through a call to flush_workqueue() and subsequently flushed once more in pci_doe_destroy_workqueue() through a call to destroy_workqueue(). Deduplicate by dropping flush_workqueue() from pci_doe_flush_mb(). Rename pci_doe_flush_mb() to pci_doe_cancel_tasks() to more aptly describe what it now does. Tested-by:
-
Lukas Wunner authored
A synchronous API for DOE has just been introduced. CXL (the only in-tree DOE user so far) was converted to use it instead of the asynchronous API. Consequently, pci_doe_submit_task() as well as the pci_doe_task struct are only used internally, so make them private. Tested-by:
-
Lukas Wunner authored
A synchronous API for DOE has just been introduced. Convert CXL CDAT retrieval over to it. Tested-by:
-
Lukas Wunner authored
The DOE API only allows asynchronous exchanges and forces callers to provide a completion callback. Yet all existing callers only perform synchronous exchanges. Upcoming commits for CMA (Component Measurement and Authentication, PCIe r6.0 sec 6.31) likewise require only synchronous DOE exchanges. Provide a synchronous pci_doe() API call which builds on the internal asynchronous machinery. Convert the internal pci_doe_discovery() to the new call. The new API allows submission of const-declared requests, necessitating the addition of a const qualifier in struct pci_doe_task. Tested-by:
Davidlohr Bueso <dave@stgolabs.net> Reviewed-by:
-
Dan Williams authored
A recent debug session yielded a couple debug messages that were useful for determining the reason why the driver was or was not falling back to CXL range register emulation, and for identifying decoder setting enumeration problems. Reviewed-by:
-
Dan Williams authored
Do not assume that a single-target port falls back to a passthrough decoder configuration. Scan for decoders and only fallback after probing that the HDM decoder capability is not present. One user visible affect of this bug is the inability to enumerate present CXL regions as the decoder settings for the present decoders are skipped. Fixes: d17d0540 ("cxl/core/hdm: Add CXL standard decoder enumeration to the core") Reported-by:
-
Dan Williams authored
After the discovery of a case where an implementation misbehaves with register reads larger than the definition of the register the other usages of readq() were audited and found to be correct, but some cases where the io-64-nonatomic-lo-hi.h include is not needed were discovered, delete them. Reviewed-by:
-
Dan Williams authored
The CXL specification mandates that 4-byte registers must be accessed with 4-byte access cycles. CXL 3.0 8.2.3 "Component Register Layout and Definition" states that the behavior is undefined if (2) 32-bit registers are accessed as an 8-byte quantity. It turns out that at least one hardware implementation is sensitive to this in practice. The @size variable results in zero with: size = readq(hdm + CXL_HDM_DECODER0_SIZE_LOW_OFFSET(which)); ...and the correct size with: lo = readl(hdm + CXL_HDM_DECODER0_SIZE_LOW_OFFSET(which)); hi = readl(hdm + CXL_HDM_DECODER0_SIZE_HIGH_OFFSET(which)); size = (hi << 32) + lo; Fixes: d17d0540 ("cxl/core/hdm: Add CXL standard decoder enumeration to the core") Cc: <stable@vger.kernel.org> Reviewed-by: -
Dan Williams authored
Decoders committed with 0-size lead to later crashes on shutdown as __cxl_dpa_release() assumes a 'struct resource' has been established in the in 'cxlds->dpa_res'. Just fail the driver load in this instance since there are deeper problems with the enumeration or the setup when this happens. Fixes: 9c57cde0 ("cxl/hdm: Enumerate allocated DPA") Cc: <stable@vger.kernel.org> Reviewed-by:
-
- 09 Apr, 2023 2 commits
-
-
Linus Torvalds authored
-
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tipLinus Torvalds authored
Pull perf fixes from Borislav Petkov: - Fix "same task" check when redirecting event output - Do not wait unconditionally for RCU on the event migration path if there are no events to migrate * tag 'perf_urgent_for_v6.3_rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: perf/core: Fix the same task check in perf_event_set_output perf: Optimize perf_pmu_migrate_context()
-
