From: "Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>
To: Ehab Ababneh <ehab.ababneh@intel.com>
Cc: Hans de Goede <hansg@kernel.org>,
tony.luck@intel.com, ashok.raj.linux@gmail.com,
platform-driver-x86@vger.kernel.org,
LKML <linux-kernel@vger.kernel.org>,
tglx@kernel.org, mingo@redhat.com, bp@alien8.de,
dave.hansen@linux.intel.com, x86@kernel.org, hpa@zytor.com,
rostedt@goodmis.org, mhiramat@kernel.org,
mathieu.desnoyers@efficios.com,
linux-trace-kernel@vger.kernel.org, peterz@infradead.org,
sathyanarayanan.kuppuswamy@linux.intel.com,
pengfei.xu@intel.com, ravi.v.shankar@intel.com
Subject: Re: [PATCH v2 09/17] platform/x86/intel/ifs: Select CPUs from topology_cluster_cpumask() for simultaneous IFS tests
Date: Thu, 19 Nov 2026 13:38:27 +0200 (EET) [thread overview]
Message-ID: <1c3ebcd7-51f8-f5fa-3bcd-75b8f2644c60@linux.intel.com> (raw)
In-Reply-To: <20261118193259.519568-10-ehab.ababneh@intel.com>
[-- Attachment #1: Type: text/plain, Size: 7718 bytes --]
On Wed, 18 Nov 2026, Ehab Ababneh wrote:
> From: Jithu Joseph <jithu.joseph@intel.com>
>
> The current intel_ifs driver selects CPUs from cpu_smt_mask() to run
> IFS test simultaneously. This approach works well for processors currently
> supported by the intel_ifs driver, where threads in a CPU core share the
> same SCAN test engine. However, it will not work for new Intel Xeon
> processors, such as Diamond Rapids, where multiple CPU cores in a module
> share the same SCAN test engine and must run IFS tests simultaneously.
>
> Starting with Granite Rapids (GNR) and later generations, bit[20:16]
> (TID_BIT_SHIFT) of MSR_INTEGRITY_CAPABILITIES register [1] contains the
> number of bits to shift right on x2APICID to obtain a unique topology ID
> for all logical processors that share a SCAN test engine. In theory,
> computing a cpumask by this method and using a stop_machine() variant
> with an arbitrary cpumask passed in to rendezvous CPUs for simultaneous
> IFS tests should work for both cases where threads in a core or CPU
> cores in a module share the same SCAN test engine. However, as per
> Tony Luck's feedback and the previous discussion [2], the Linux community
> is not in favor of a stop_machine() variant with an arbitrary cpumask.
I suggest using passive voice for what comes after ,
>
> Observations on existing Intel Xeon processors and the upcoming Diamond
> Rapids with IFS capability show that the logical processors sharing the
> same SCAN test engine are exactly the logical processors sharing the
> same L2 cache.
>
> topology_cluster_cpumask() returns the cpumask for CPUs sharing the same
> L2 cache. Therefore, select CPUs from topology_cluster_cpumask() to run
> IFS tests if simultaneous testing is required.
>
> [ Qiuxu: Re-wrote the commit message. ]
There's no Signed-off from Qiuxu ?
> [1] Vol. 4 "Table 2-56. Additional MSRs Supported by the Intel® Xeon®
> 6 P-Core Processors" in the Intel Software Developers' Manual,
> March 2025.
>
> [2] https://lore.kernel.org/all/20220505082824.GD2501@worktop.programming.kicks-ass.net/
Move to Link: and add [2] to end.
>
> Suggested-by: Tony Luck <tony.luck@intel.com>
> Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
> Signed-off-by: Ehab Ababneh <ehab.ababneh@intel.com>
> Tested-by: Yi Lai <yi1.lai@intel.com>
> ---
> drivers/platform/x86/intel/ifs/ifs.h | 2 +
> drivers/platform/x86/intel/ifs/runtest.c | 54 ++++++++++++++++++------
> 2 files changed, 44 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/platform/x86/intel/ifs/ifs.h b/drivers/platform/x86/intel/ifs/ifs.h
> index 814ebac10a5d..365429fafe7a 100644
> --- a/drivers/platform/x86/intel/ifs/ifs.h
> +++ b/drivers/platform/x86/intel/ifs/ifs.h
> @@ -285,6 +285,7 @@ struct ifs_test_msrs {
>
> /**
> * struct ifs_data - attributes related to Intel IFS driver
> + * @grp_cpumask: set of CPUs being tested simultaneously
> * @loaded_version: stores the currently loaded ifs image version.
> * @loaded: If a valid test binary has been loaded into the memory
> * @loading_error: Error occurred on another CPU while loading image
> @@ -298,6 +299,7 @@ struct ifs_test_msrs {
> * @all_lp_join: all logical processors join test
> */
> struct ifs_data {
> + cpumask_t grp_cpumask;
> int loaded_version;
> bool loaded;
> bool loading_error;
> diff --git a/drivers/platform/x86/intel/ifs/runtest.c b/drivers/platform/x86/intel/ifs/runtest.c
> index c6a4d6dcccf1..c618857cd620 100644
> --- a/drivers/platform/x86/intel/ifs/runtest.c
> +++ b/drivers/platform/x86/intel/ifs/runtest.c
> @@ -156,11 +156,9 @@ static atomic_t scan_cpus_in;
> /*
> * Simplified CPU sibling rendezvous loop based on microcode loader wait_for_cpus()
> */
> -static void wait_for_sibling_cpu(atomic_t *t, long long timeout)
> +static void wait_for_sibling_cpu(struct ifs_data *ifsd, atomic_t *t, long long timeout)
> {
> - int cpu = smp_processor_id();
> - const struct cpumask *smt_mask = cpu_smt_mask(cpu);
> - int all_cpus = cpumask_weight(smt_mask);
> + int all_cpus = cpumask_weight(&ifsd->grp_cpumask);
>
> atomic_inc(t);
> while (atomic_read(t) < all_cpus) {
> @@ -195,9 +193,9 @@ static int doscan(void *data)
> }
>
> /* Only the first logical CPU on a core reports result */
> - first = cpumask_first(cpu_smt_mask(cpu));
> + first = cpumask_first(&ifsd->grp_cpumask);
>
> - wait_for_sibling_cpu(&scan_cpus_in, NSEC_PER_SEC);
> + wait_for_sibling_cpu(ifsd, &scan_cpus_in, NSEC_PER_SEC);
>
> /*
> * This WRMSR will wait for other HT threads to also write
> @@ -265,7 +263,11 @@ static void ifs_test_core(int cpu, struct device *dev)
>
> params.activate = &activate;
> atomic_set(&scan_cpus_in, 0);
> - stop_core_cpuslocked(cpu, doscan, ¶ms);
> +
> + if (ifsd->all_lp_join)
> + stop_cluster_cpuslocked(cpu, doscan, ¶ms);
> + else
> + stop_core_cpuslocked(cpu, doscan, ¶ms);
>
> status = params.status;
>
> @@ -310,15 +312,18 @@ static int do_array_test(void *data)
> struct run_array_params *params = data;
> int cpu = smp_processor_id();
> union ifs_array *command;
> + struct ifs_data *ifsd;
> int first;
>
> + ifsd = params->ifsd;
> +
> command = params->command;
> - wait_for_sibling_cpu(&array_cpus_in, NSEC_PER_SEC);
> + wait_for_sibling_cpu(ifsd, &array_cpus_in, NSEC_PER_SEC);
>
> /*
> * Only one logical CPU on a core needs to trigger the Array test via MSR write.
> */
> - first = cpumask_first(cpu_smt_mask(cpu));
> + first = cpumask_first(&ifsd->grp_cpumask);
>
> if (cpu == first) {
> wrmsrq(MSR_ARRAY_BIST, command->data);
> @@ -350,7 +355,11 @@ static void ifs_array_test_core(int cpu, struct device *dev)
> }
> atomic_set(&array_cpus_in, 0);
> params.command = &command;
> - stop_core_cpuslocked(cpu, do_array_test, ¶ms);
> +
> + if (ifsd->all_lp_join)
> + stop_cluster_cpuslocked(cpu, do_array_test, ¶ms);
> + else
> + stop_core_cpuslocked(cpu, do_array_test, ¶ms);
Add a helper for this or should it just be a pointer (or add ops)?
>
> if (command.ctrl_result)
> break;
> @@ -398,6 +407,24 @@ static void ifs_array_test_gen1(int cpu, struct device *dev)
> ifsd->status = SCAN_TEST_PASS;
> }
>
> +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
> @@ -420,12 +447,15 @@ int do_core_test(int cpu, struct device *dev)
>
> switch (test->test_num) {
> case IFS_TYPE_SAF:
> - if (!ifsd->loaded)
> + if (!ifsd->loaded) {
> ret = -EPERM;
> - else
> + } else {
> + build_cpugroup_mask(dev, cpu);
> ifs_test_core(cpu, dev);
> + }
> break;
> case IFS_TYPE_ARRAY_BIST:
> + build_cpugroup_mask(dev, cpu);
> if (ifsd->array_gen == ARRAY_GEN0)
> ifs_array_test_core(cpu, dev);
> else
>
Unrelated to this patch, this function would be better of with using
guard() + simplied code flow it allows.
--
i.
next prev parent reply other threads:[~2026-11-19 11:38 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-18 19:32 [PATCH v2 00/17] platform/x86/intel/ifs: Support group-based In-Field Scan and Diamond Rapids Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 01/17] platform/x86/intel/ifs: Revert SBAF feature Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 02/17] platform/x86/intel/ifs: Fix IFS sysfs interfaces in the comments Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 03/17] platform/x86/intel/ifs: Clarify In-Field Scan MSR bit macro definitions Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 04/17] platform/x86/intel/ifs: Use u64 to define the bits of IFS MSR registers Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 05/17] platform/x86/intel/ifs: Early return on the IFS image hash/chunk error Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 06/17] platform/x86/intel/ifs: Refactor Array BIST function interface Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 07/17] stop_machine: Add stop_cluster_cpuslocked() for per-cluster operations Ehab Ababneh
2026-11-19 11:28 ` Ilpo Järvinen
2026-11-18 19:32 ` [PATCH v2 08/17] platform/x86/intel/ifs: Enumerate the bit to determine if CPUs need to join for IFS tests Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 09/17] platform/x86/intel/ifs: Select CPUs from topology_cluster_cpumask() for simultaneous " Ehab Ababneh
2026-11-19 11:38 ` Ilpo Järvinen [this message]
2026-11-18 19:32 ` [PATCH v2 10/17] platform/x86/intel/ifs: Warn if a group test fails with CPUs missing Ehab Ababneh
2026-11-19 11:42 ` Ilpo Järvinen
2026-11-18 19:32 ` [PATCH v2 11/17] platform/x86/intel/ifs: Track and report scan results on a per-CPU basis Ehab Ababneh
2026-11-18 19:47 ` sashiko-bot
2026-11-18 19:32 ` [PATCH v2 12/17] platform/x86/intel/ifs: Add new sysfs interfaces {status, details, cpus_ran}_list Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 13/17] platform/x86/intel/ifs: Add additional details sysfs interface Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 14/17] platform/x86/intel/ifs: Add Intel Diamond Rapids server support Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 15/17] platform/x86/intel/ifs: Remove limit on number of batch files Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 16/17] platform/x86/intel/ifs: Update IFS documentation Ehab Ababneh
2026-11-18 19:32 ` [PATCH v2 17/17] Documentation/ABI: Update IFS ABI doc Ehab Ababneh
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1c3ebcd7-51f8-f5fa-3bcd-75b8f2644c60@linux.intel.com \
--to=ilpo.jarvinen@linux.intel.com \
--cc=ashok.raj.linux@gmail.com \
--cc=bp@alien8.de \
--cc=dave.hansen@linux.intel.com \
--cc=ehab.ababneh@intel.com \
--cc=hansg@kernel.org \
--cc=hpa@zytor.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-trace-kernel@vger.kernel.org \
--cc=mathieu.desnoyers@efficios.com \
--cc=mhiramat@kernel.org \
--cc=mingo@redhat.com \
--cc=pengfei.xu@intel.com \
--cc=peterz@infradead.org \
--cc=platform-driver-x86@vger.kernel.org \
--cc=ravi.v.shankar@intel.com \
--cc=rostedt@goodmis.org \
--cc=sathyanarayanan.kuppuswamy@linux.intel.com \
--cc=tglx@kernel.org \
--cc=tony.luck@intel.com \
--cc=x86@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox