Re: [PATCH v23 4/8] crash: memory and CPU hotplug sysfs attributes
From: Eric DeVolder
Date: Tue Jun 13 2023 - 15:59:38 EST
On 6/13/23 10:24, Eric DeVolder wrote:
On 6/13/23 03:03, Greg KH wrote:
On Mon, Jun 12, 2023 at 05:07:08PM -0400, Eric DeVolder wrote:
Introduce the crash_hotplug attribute for memory and CPUs for
use by userspace. These attributes directly facilitate the udev
rule for managing userspace re-loading of the crash kernel upon
hot un/plug changes.
For memory, expose the crash_hotplug attribute to the
/sys/devices/system/memory directory. For example:
# udevadm info --attribute-walk /sys/devices/system/memory/memory81
looking at device '/devices/system/memory/memory81':
KERNEL=="memory81"
SUBSYSTEM=="memory"
DRIVER==""
ATTR{online}=="1"
ATTR{phys_device}=="0"
ATTR{phys_index}=="00000051"
ATTR{removable}=="1"
ATTR{state}=="online"
ATTR{valid_zones}=="Movable"
looking at parent device '/devices/system/memory':
KERNELS=="memory"
SUBSYSTEMS==""
DRIVERS==""
ATTRS{auto_online_blocks}=="offline"
ATTRS{block_size_bytes}=="8000000"
ATTRS{crash_hotplug}=="1"
For CPUs, expose the crash_hotplug attribute to the
/sys/devices/system/cpu directory. For example:
# udevadm info --attribute-walk /sys/devices/system/cpu/cpu0
looking at device '/devices/system/cpu/cpu0':
KERNEL=="cpu0"
SUBSYSTEM=="cpu"
DRIVER=="processor"
ATTR{crash_notes}=="277c38600"
ATTR{crash_notes_size}=="368"
ATTR{online}=="1"
looking at parent device '/devices/system/cpu':
KERNELS=="cpu"
SUBSYSTEMS==""
DRIVERS==""
ATTRS{crash_hotplug}=="1"
ATTRS{isolated}==""
ATTRS{kernel_max}=="8191"
ATTRS{nohz_full}==" (null)"
ATTRS{offline}=="4-7"
ATTRS{online}=="0-3"
ATTRS{possible}=="0-7"
ATTRS{present}=="0-3"
With these sysfs attributes in place, it is possible to efficiently
instruct the udev rule to skip crash kernel reloading for kernels
configured with crash hotplug support.
For example, the following is the proposed udev rule change for RHEL
system 98-kexec.rules (as the first lines of the rule file):
# The kernel updates the crash elfcorehdr for CPU and memory changes
SUBSYSTEM=="cpu", ATTRS{crash_hotplug}=="1", GOTO="kdump_reload_end"
SUBSYSTEM=="memory", ATTRS{crash_hotplug}=="1", GOTO="kdump_reload_end"
When examined in the context of 98-kexec.rules, the above rules
test if crash_hotplug is set, and if so, the userspace initiated
unload-then-reload of the crash kernel is skipped.
CPU and memory checks are separated in accordance with
CONFIG_HOTPLUG_CPU and CONFIG_MEMORY_HOTPLUG kernel config options.
If an architecture supports, for example, memory hotplug but not
CPU hotplug, then the /sys/devices/system/memory/crash_hotplug
attribute file is present, but the /sys/devices/system/cpu/crash_hotplug
attribute file will NOT be present. Thus the udev rule skips
userspace processing of memory hot un/plug events, but the udev
rule will evaluate false for CPU events, thus allowing userspace to
process CPU hot un/plug events (ie the unload-then-reload of the kdump
capture kernel).
Signed-off-by: Eric DeVolder <eric.devolder@xxxxxxxxxx>
Reviewed-by: Sourabh Jain <sourabhjain@xxxxxxxxxxxxx>
Acked-by: Hari Bathini <hbathini@xxxxxxxxxxxxx>
Acked-by: Baoquan He <bhe@xxxxxxxxxx>
---
.../admin-guide/mm/memory-hotplug.rst | 8 ++++++++
Documentation/core-api/cpu_hotplug.rst | 18 ++++++++++++++++++
drivers/base/cpu.c | 14 ++++++++++++++
drivers/base/memory.c | 13 +++++++++++++
include/linux/kexec.h | 8 ++++++++
5 files changed, 61 insertions(+)
diff --git a/Documentation/admin-guide/mm/memory-hotplug.rst
b/Documentation/admin-guide/mm/memory-hotplug.rst
index 1b02fe5807cc..eb99d79223a3 100644
--- a/Documentation/admin-guide/mm/memory-hotplug.rst
+++ b/Documentation/admin-guide/mm/memory-hotplug.rst
@@ -291,6 +291,14 @@ The following files are currently defined:
Availability depends on the CONFIG_ARCH_MEMORY_PROBE
kernel configuration option.
``uevent`` read-write: generic udev file for device subsystems.
+``crash_hotplug`` read-only: when changes to the system memory map
+ occur due to hot un/plug of memory, this file contains
+ '1' if the kernel updates the kdump capture kernel memory
+ map itself (via elfcorehdr), or '0' if userspace must update
+ the kdump capture kernel memory map.
+
+ Availability depends on the CONFIG_MEMORY_HOTPLUG kernel
+ configuration option.
====================== =========================================================
.. note::
diff --git a/Documentation/core-api/cpu_hotplug.rst b/Documentation/core-api/cpu_hotplug.rst
index f75778d37488..0c8dc3fe5f94 100644
--- a/Documentation/core-api/cpu_hotplug.rst
+++ b/Documentation/core-api/cpu_hotplug.rst
@@ -750,6 +750,24 @@ will receive all events. A script like::
can process the event further.
+When changes to the CPUs in the system occur, the sysfs file
+/sys/devices/system/cpu/crash_hotplug contains '1' if the kernel
+updates the kdump capture kernel list of CPUs itself (via elfcorehdr),
+or '0' if userspace must update the kdump capture kernel list of CPUs.
+
+The availability depends on the CONFIG_HOTPLUG_CPU kernel configuration
+option.
+
+To skip userspace processing of CPU hot un/plug events for kdump
+(ie the unload-then-reload to obtain a current list of CPUs), this sysfs
+file can be used in a udev rule as follows:
+
+ SUBSYSTEM=="cpu", ATTRS{crash_hotplug}=="1", GOTO="kdump_reload_end"
+
+For a cpu hot un/plug event, if the architecture supports kernel updates
+of the elfcorehdr (which contains the list of CPUs), then the rule skips
+the unload-then-reload of the kdump capture kernel.
+
Kernel Inline Documentations Reference
======================================
diff --git a/drivers/base/cpu.c b/drivers/base/cpu.c
index c1815b9dae68..06a0c22b37b8 100644
--- a/drivers/base/cpu.c
+++ b/drivers/base/cpu.c
@@ -282,6 +282,17 @@ static ssize_t print_cpus_nohz_full(struct device *dev,
static DEVICE_ATTR(nohz_full, 0444, print_cpus_nohz_full, NULL);
#endif
+#ifdef CONFIG_HOTPLUG_CPU
+#include <linux/kexec.h>
+static ssize_t crash_hotplug_show(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ return sprintf(buf, "%d\n", crash_hotplug_cpu_support());
+}
+static DEVICE_ATTR_ADMIN_RO(crash_hotplug);
+#endif
+
static void cpu_device_release(struct device *dev)
{
/*
@@ -469,6 +480,9 @@ static struct attribute *cpu_root_attrs[] = {
#ifdef CONFIG_NO_HZ_FULL
&dev_attr_nohz_full.attr,
#endif
+#ifdef CONFIG_HOTPLUG_CPU
+ &dev_attr_crash_hotplug.attr,
+#endif
#ifdef CONFIG_GENERIC_CPU_AUTOPROBE
&dev_attr_modalias.attr,
#endif
diff --git a/drivers/base/memory.c b/drivers/base/memory.c
index b456ac213610..24b8ef4c830c 100644
--- a/drivers/base/memory.c
+++ b/drivers/base/memory.c
@@ -490,6 +490,16 @@ static ssize_t auto_online_blocks_store(struct device *dev,
static DEVICE_ATTR_RW(auto_online_blocks);
+#ifdef CONFIG_MEMORY_HOTPLUG
+#include <linux/kexec.h>
+static ssize_t crash_hotplug_show(struct device *dev,
+ struct device_attribute *attr, char *buf)
+{
+ return sprintf(buf, "%d\n", crash_hotplug_memory_support());
+}
This sysfs file has to be documented in Documentation/ABI/ right?
I will add these entries.
And did you use checkpatch? It should have told you to use sysfs_emit()
instead...
I did use checkpatch, and it did not reveal that.
[root@localhost linux]# ./scripts/checkpatch.pl --patch
../v23/0004-crash-memory-and-CPU-hotplug-sysfs-attributes.patch
total: 0 errors, 0 warnings, 103 lines checked
../v23/0004-crash-memory-and-CPU-hotplug-sysfs-attributes.patch has no obvious style problems and is
ready for submission.
This is not the first time it's been suggested to me that checkpatch should have caught a problem;
is there a different invocation than what I'm doing?
+static DEVICE_ATTR_RO(crash_hotplug);
+#endif
All of these #ifdefs should all be removed and instead use the
is_visible() callback to determine if the attribute is shown or not,
using the IS_ENABLED() test in the function.
ok, I'll correct this.
I've been examining drivers/base/cacheinfo.c as a template for how to remove the
#ifdefs and use the is_visible() callback for the drivers/base/cpu|memory.c files.
I'm attempting to apply this technique to drivers/base/cpu.c. In this file, there
are features that are compiled in/out based on the CONFIG settings, for example
CONFIG_ARCH_CPU_PROBE_RELEASE. My attempts at applying the technique thus far have
resulted in link-time errors for missing symbols, ie. arch_cpu_probe() and
arch_cpu_release().
As I understand it, to use IS_ENABLED(XYZ) to guard-band conditional code, the contents
of that code still needs to be compile-able (eg. no references to struct members with
surrounding #ifdef CONFIG_XYZ) and link-able (eg. any called functions must also be
compiled).
Given my understanding of the request and IS_ENABLED, I'm not seeing a path forward.
I'm thinking the approach works well in cacheinfo.c for example because cache info is
always compiled-in.
I suspect I am misunderstanding, and would appreciate a pointer/guidance.
Thanks!
eric
Thank you for looking at this!
eric
thanks,
greg k-h