Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider: - [High] platform/x86/intel/ifs: ifs_init() fails on SAF_GEN=0 CPUs - [High] platform/x86/intel/ifs: do_core_test() updates grp_cpumask before validation - [Low] platform/x86/intel/ifs: fix documentation typo and inconsistencies in ifs.h --
commit 00e0613f0da8feb277c2b0b17ada7d9df4902b12 Author: Qiuxu Zhuo <[email protected]> platform/x86/intel/ifs: Update IFS documentation This series updates the Intel In-Field Scan (IFS) driver to support group-based simultaneous testing using topology_cluster_cpumask(), validates TID_BIT_SHIFT against the cluster topology, and adds per-group sysfs result attributes and documentation. > diff --git a/drivers/platform/x86/intel/ifs/core.c > b/drivers/platform/x86/intel/ifs/core.c > index 3dafa28a10104..5c9c86e048371 100644 > --- a/drivers/platform/x86/intel/ifs/core.c > +++ b/drivers/platform/x86/intel/ifs/core.c [ ... ] > +/* > + * Test whether logical processors indicated by TID_BIT_SHIFT that share the > + * same SCAN test engine match the cpumask from topology_cluster_cpumask(). > + */ > +static bool topology_cluster_matched(u32 msrval) > +{ > + u32 dist, shift = FIELD_GET(MSR_INTEGRITY_CAPS_TID_BIT_SHIFT_MASK, > msrval); > + const cpumask_t *cpumask = topology_cluster_cpumask(0); > + u32 delta, step = 0, first = cpu_data(0).topo.apicid; > + unsigned int cpu, cpus; > + > + /* Get the minimal APIC ID step among the CPUs in the same cluster. */ > + for_each_cpu_andnot(cpu, cpumask, cpumask_of(0)) { > + delta = abs(cpu_data(cpu).topo.apicid - first); > + > + if (!step) { > + step = delta; > + continue; > + } > + > + if (step > delta) > + step = delta; > + } > + > + /* A Hyper-Threading (HT) system with HT disabled. */ > + if (!step) > + step = 2; > + > + cpus = cpumask_weight(cpumask); > + dist = step * cpus; > + > + if (dist != (1 << shift)) { > + pr_err_once("APICID step %u, group cpus %u, TID_BIT_SHIFT > %u\n", step, cpus, shift); > + return false; > + } > + > + return true; > +} [ ... ] > @@ -134,11 +157,33 @@ static int __init ifs_init(void) > continue; > ifs_devices[i].rw_data.generation = > FIELD_GET(MSR_INTEGRITY_CAPS_SAF_GEN_MASK, > msrval); > + /* > + * For Sapphire Rapids and Emerald Rapids (SAF_GEN=0), the > ALL_LP_JOIN > + * indicator is unavailable. However, all the threads within a > core still > + * need to run IFS simultaneously, so the variable > 'all_lp_join' is > + * forced to true. > + * > + * For Granite Rapids and later generations (SAF_GEN=2), the > ALL_LP_JOIN > + * indicator is available, the variable 'all_lp_join' is > determined by > + * this indicator. > + */ > + ifs_devices[i].rw_data.all_lp_join = > ifs_devices[i].rw_data.generation ? > + (msrval & > MSR_INTEGRITY_CAPS_ALL_LP_JOIN) : true; > + > + if (ifs_devices[i].rw_data.all_lp_join && > !topology_cluster_matched(msrval)) { > + pr_err_once("TID_BIT_SHIFT mismatched > topology_cluster_cpumask().\n"); > + ret = -EINVAL; > + goto err_exit; > + } [Severity: High] Does this cause ifs_init() to fail with -EINVAL on all generation == 0 platforms and on systems where any CPU in cluster 0 is offline at module load time? In ifs_init(), when ifs_devices[i].rw_data.generation is 0 (such as on Sapphire Rapids, Emerald Rapids, Sierra Forest, and Clearwater Forest), all_lp_join is forced to true and topology_cluster_matched(msrval) is called unconditionally. However, TID_BIT_SHIFT (MSR_INTEGRITY_CAPS_TID_BIT_SHIFT_MASK) was only introduced starting with Granite Rapids (generation == 2) and is 0 on generation == 0 processors. In topology_cluster_matched(): cpus = cpumask_weight(cpumask); dist = step * cpus; if (dist != (1 << shift)) { ... return false; } Because step is at least 1 when siblings are present (with cpus >= 2) or set to 2 when !step (with cpus == 1), dist is always >= 2, whereas (1 << shift) is 1 when shift is 0. As a result, topology_cluster_matched() always returns false on generation == 0 CPUs, aborting driver initialization with -EINVAL (and on Sierra Forest and Clearwater Forest, forcing all_lp_join to true also causes build_cpugroup_mask() to use the 4-core L2 cluster mask instead of a single core). In addition, topology_cluster_matched() reads topology_cluster_cpumask(0) without holding cpus_read_lock() and assumes all logical processors in CPU 0's cluster are online. Since cpu_l2c_shared_map only tracks online CPUs, if any sibling thread or core in cluster 0 is offline when the module loads (for example, 3 of 4 CPUs online in a Dual-Core Module giving step = 1 and cpus = 3, or only CPU 0 online giving step = 2 and cpus = 1 when shift = 2), dist != (1 << shift) will also evaluate to true and fail initialization. > diff --git a/drivers/platform/x86/intel/ifs/ifs.h > b/drivers/platform/x86/intel/ifs/ifs.h > index f369fb0d3d82f..551202f86cea4 100644 > --- a/drivers/platform/x86/intel/ifs/ifs.h > +++ b/drivers/platform/x86/intel/ifs/ifs.h > @@ -81,30 +81,52 @@ > * The IFS driver provides sysfs interfaces via > /sys/devices/virtual/misc/intel_ifs_<n>/ > * to control execution: > * > - * Test a specific core:: > + * Test a specific group:: > * > - * # echo <cpu#> > /sys/devices/virtual/misc/intel_ifs_<n>/run_test > + * # echo <CPU#> > /sys/devices/virtual/misc/intel_ifs_<n>/run_test > * > - * when HT is enabled any of the sibling cpu# can be specified to test > - * its corresponding physical core. Since the tests are per physical core, > - * the result of testing any thread is same. All siblings must be online > - * to run a core test. It is only necessary to test one thread. > + * Any CPU in the same group that shares the same SCAN test engin can be > + * specified to test its corresponding group. All CPUs in the same group > + * must be online to run a group test. It is only necessary to test one > + * CPU in the group. [ ... ] > + * More additional details of the last test for all CPUs in the same group > are provided in /sys:: > + * > + * $ cat /sys/devices/virtual/misc/intel_ifs_0/addnl_details_list > + * 0xc44191a 0xc44191a > * > * The details file reports the hex value of the test specific status MSR. > * Hardware defined error codes are documented in volume 4 of the Intel > @@ -127,48 +149,22 @@ > * 2) Hardware allows for some number of cores to be tested in parallel. > * The driver does not make use of this, it only tests one core at a time. [Severity: Low] This isn't a runtime bug, but there are a few small inconsistencies in the documentation update in drivers/platform/x86/intel/ifs/ifs.h: 1. "shares the same SCAN test engin can be" has a typo ("engin" instead of "engine"). 2. Under "Driver design choices", item 2 still says "it only tests one core at a time" rather than one group at a time. 3. The commit message refers to "{status, details, additional_details, cpus_ran}_list", whereas the sysfs attribute documented here and implemented in drivers/platform/x86/intel/ifs/sysfs.c is named addnl_details_list. > diff --git a/drivers/platform/x86/intel/ifs/runtest.c > b/drivers/platform/x86/intel/ifs/runtest.c > index a882cd42c191a..c57b9ed0b7be7 100644 > --- a/drivers/platform/x86/intel/ifs/runtest.c > +++ b/drivers/platform/x86/intel/ifs/runtest.c [ ... ] > +static void build_cpugroup_mask(struct device *dev, int cpu) > +{ > + struct ifs_data *ifsd = ifs_get_data(dev); > + > + cpumask_clear(&ifsd->grp_cpumask); > + > + /* > + * Some CPUs, such as Sierra Forest and Clearwater Forest, have > multiple cores > + * sharing the same SCAN test engine, but they don't require a > rendezvous. > + */ > + if (!ifsd->all_lp_join) { > + cpumask_set_cpu(cpu, &ifsd->grp_cpumask); > + return; > + } > + > + cpumask_copy(&ifsd->grp_cpumask, topology_cluster_cpumask(cpu)); > +} > + > /* > - * Initiate per core test. It wakes up work queue threads on the target cpu > and > - * its sibling cpu. Once all sibling threads wake up, the scan test gets > executed and > + * Initiate per core test. It wakes up work queue threads on the target CPU > and > + * its sibling CPU. Once all sibling threads wake up, the scan test gets > executed and > * wait for all sibling threads to finish the scan test. > */ > int do_core_test(int cpu, struct device *dev) > @@ -628,9 +493,10 @@ int do_core_test(int cpu, struct device *dev) > > /* Prevent CPUs from being taken offline during the scan test */ > cpus_read_lock(); > + build_cpugroup_mask(dev, cpu); > > if (!cpu_online(cpu)) { > - dev_info(dev, "cannot test on the offline cpu %d\n", cpu); > + dev_info(dev, "cannot test on the offline CPU %d\n", cpu); > ret = -EINVAL; > goto out; > } [Severity: High] Should build_cpugroup_mask(dev, cpu) be called only after verifying that cpu is online and that a test will actually be executed? In run_test_store() -> do_core_test(), run_test_store() only checks cpu >= nr_cpu_ids before calling do_core_test(cpu, dev), and do_core_test() calls build_cpugroup_mask(dev, cpu) before checking !cpu_online(cpu) or !ifsd->loaded: 1. When ifsd->all_lp_join is true, build_cpugroup_mask() calls: cpumask_copy(&ifsd->grp_cpumask, topology_cluster_cpumask(cpu)); which accesses per_cpu(cpu_l2c_shared_map, cpu). Because smp_prepare_cpus_common() only allocates cpu_l2c_shared_map for CPUs in cpu_possible_mask, writing a CPU number with cpu < nr_cpu_ids that is not in cpu_possible_mask when CONFIG_CPUMASK_OFFSTACK=y will dereference a NULL cpumask pointer before the !cpu_online(cpu) check is reached. 2. If do_core_test() returns early with -EINVAL (because cpu is offline) or with -EPERM (because !ifsd->loaded for IFS_TYPE_SAF), build_cpugroup_mask() has already overwritten ifsd->grp_cpumask. Subsequent reads of cpus_ran_list, status_list, details_list, and addnl_details_list in drivers/platform/x86/intel/ifs/sysfs.c will then report the CPUs and stale per-CPU state of the rejected request instead of "none" or the last test that actually ran. -- Sashiko AI review ยท https://sashiko.dev/#/patchset/[email protected]?part=16
